In what is becoming an annual tradition, we are “Live from Nvidia GTC25” AI-everywhere show. We cover everything from industry landscape to Hopper to Blackwell to Rubin and Feynman, plus silicon photonics for cluster interconnect fabric (star of the show, really), the complexity of inference for customers, low-end systems, power and cooling (did we hear 600 KW per rack?), software including cluster-level AI-workload-focused open-sourced Dynamo “OS”, and storage (the semantic kind).
Just before the GTC25 conference, and in the 100th episode of the full format @HPCpodcast, we welcome a very special guest, the great Dr. Ian Cutress, Chief Analyst at More Than Moore and host of the popular video channel TechTechPotato to discuss the state of AI and advanced chips, new technologies and architectures, the startup scene, and top trends in semiconductor design & manufacturing. Join us!
We are delighted to have as special guests today three of the top analysts in the HPC, AI, Cloud, and Quantum fields, representing the industry analyst firm Hyperion Research.Earl Joseph is Hyperion CEO with oversight for Hyperion’s overall research and consulting efforts along with the firm’s HPC User Forum, which conducts conferences in the US and around the world throughout the year, Mark Nossokoff, Research Director and Lead Analyst for Storage and Interconnects and Cloud, and Bob Sorensen, Senior VP of Research, who focuses on Exascale, Supercomputing, Quantum, and other future technologies.
Join us for an In Depth discussion of the current state and future trends in HPC, AI, Quantum, Cloud Computing, Exascale, Storage, Interconnects and Optical I/O, and Liquid Cooling.
In this In-Depth feature of the @HPDpodcast, Addison Snell, co-founder and CEO of Intersect360 joins Shahin and Doug as they discuss a wide range of topics in HPC, AI, and Quantum Computing. They touch on HPC-AI market size, the impact of hyperscalers on AI, the future of leadership computing facilities (exascale.next), AI nationalism, DeepSeek AI, Nvidia’s leadership and challenges, the state of quantum computing, AI storage, optical computing and interconnects, and of course, Addison’s crossword puzzles features in The New York Times and Wall Street Journal! Make sure, especially, to look up the crossword puzzle in the Sunday Times of October, 30th, 2022.
This is a great time to review what we learned at some of the key conferences that were held towards the end of last year. Adrian Cockcroft joins Shahin Khan and Doug Black to discuss SC24, RISC-V Summit, and AWS Reinvent. Topics include: HPC and AI Clouds, CXL, Liquid Cooling, Optical Interconnects, Optical Computing, Novel CPUs and GPUs, the state of RISC-V in servers and supercomputers, TOP500, Chiplets, AWS CPU and GPU strategies. We recorded this episode just before the always-good Q2B conference on quantum technologies, which is mentioned here and will be covered in a future episode.
Special guest and a leading light of the HPC industry, Prof. Torsten Hoefler of ETH-Zurich joins Shahin and Doug in a lively discussion about the Age of Computation, Ultra Ethernet, datacenter power and cooling, the creative process for AI, model certainty for AI, AI and emergent behavior, and other HPC topics.
Torsten is a professor at ETH Zurich, where he directs the Scalable Parallel Computing Laboratory. He is also the Chief Architect for machine learning at the Swiss National Supercomputing Center, as well as a consultant for Microsoft on large scale AI and networking.
Special guest David Kanter of ML Commons joins Shahin Khan and Doug Black to discuss AI performance metrics. AI is everywhere and destined to run on everything from devices to big systems. So, in addition to the well-known MLPerf benchmark for AI training, ML Commons provides a growing suite of benchmarks and data sets for other aspects of AI such as inference, storage, and safety. David is a founder, board member of ML Commons and the head of MLPerf benchmarks. David has more than 16 years of experience in semiconductors computing and machine learning. He was a founder of a microprocessor and compiler startup. He was also at Astor Data Systems and has consulted for NVIDIA, Intel, KLA, Applied Materials, Qualcomm, Microsoft, and others.
Last year about this time, we had the opportunity to discuss the state of HPC and the Aurora supercomputer with Rick Stevens and Mike Papka of Argonne National Lab. In the run up to SC24, we are delighted to do the same! Rick and Mike kindly carved out some time to join us for another wide ranging discussion.
We discuss Aurora, Exascale, AI, reliability at scale, technology adoption agility, datacenter power and cooling, cloud computing, quantum computing.
We’d like to encourage you to also listen to episodes 15 and 16 where we discuss AI in science with Prof. Stevens, and epsideo 75 referenced above, just before SC23.
Rick Stevens is Argonne’s Associate Laboratory Director for the Computing, Environment and Life Sciences (CELS) Directorate and an Argonne Distinguished Fellow. He is also a Professor of Computer Science at the University of Chicago. He was previously leader of Exascale Computing Initiative at Argonne.
Michael Papka is a senior scientist at Argonne National Laboratory where he is also deputy associate laboratory director for Computing, Environment and Life Sciences (CELS) and division director of the Argonne Leadership Computing Facility (ALCF).
The @HPCpodcasts Industry View feature takes on major issues in the world of HPC, AI, and other advanced technologies through the lens of industry leaders. Today, we discuss the design and deployment of large scale AI infrastructure: why AI at scale is such a critical need, where the challenges lie, and what it takes to do it right. We are joined by Jonathan Ha who is Senior Director of Product Management for AI at Penguin Solutions. Jonathan has been in the industry for more than 25 years has previously held senior positions in product management at Microsoft, AMD, and AWS.
Special guest Mahta Emrani joins us to discuss the hot topic of AI and synthetic data for market research:
– AI will usher in fast, cheap, high quality research, right?
– It’s all about details and nuances
– Why do research?
Shahin and Doug are joined by Nestor Maslej of the Stanford Institute for Human-Centered Artificial Intelligence (HAI) at Stanford University. He tracks the advancement of AI in his role as Research Manager and Editor in Chief of the annual Stanford AI Index and Stanford Global AI Vibrancy Tool. Nestor has degrees from Harvard and Oxford and is also a fellow at the Center for International Governance Innovation.
A 502 page report, the AI Index covers a wide range of topics in nine chapters:
1) Research and Development
2)Technical Performance
3) Responsible AI
4) Economy
5) Science and Medicine
6) Education
7) Policy and Governance
8) Diversity
9) Public Opinion
Special guest Sam Brealey joins us as we discuss marketing strategy and execution, specially for small businesses, “pivot to video, and AI in marketing. Cartoon of the week on marketing (or is it sales?) strategy kicks off the discussion. Join us!
2023 Year in Review is our annual special edition as we look back at one of the more eventful years in recent history for HPC, AI, Quantum Computing, and other advanced technologies. The list below includes time stamps (in minutes and seconds) and the associated topic in the podcast.
02:00 – HPC
03:45 – AI
08:03 – Metaverse
12:01 – Chips, GPUs, Accelerators 14:00 – GPU Competition
14:00 – GPU Competition
15:46 – Open Source
17:54 – Aurora Supercomputer 20:21 – TOP500 20:55 – Cloud in TOP10 21:53 – China
24:15 – Europe
25:55 – Quantum Computing
30:12 – Photonics
31:35 – Cryptocurrencies
As SC23 approaches, we were fortunate to catch up with Rick Stevens and Mike Papka of Argonne National Lab for a wide ranging discussion. In addition to an update on the Aurora supercomputer and TOP500, we also discuss the need and challenged of building a national exascale capability, developing teams and bench strength, the risks and opportunities of AI for science and society, the trend towards integrated research infrastructure (IRI), and what’s next for the exascale initiative. We’d like to encourage you to also listen to episodes 15 and 16 of this podcast where we discuss AI in science with prof. Stevens.
Rick Stevens is Argonne’s Associate Laboratory Director for the Computing, Environment and Life Sciences (CELS) Directorate and an Argonne Distinguished Fellow. He is also a Professor of Computer Science at the University of Chicago. He was previously leader of Exascale Computing Initiative at Argonne.
Michael Papka is a senior scientist at Argonne National Laboratory where he is also deputy associate laboratory director for Computing, Environment and Life Sciences (CELS) and division director of the Argonne Leadership Computing Facility (ALCF).
Karl Freund, founder and principal analyst at Cambrian-AI Research joins us to discuss the, well, “Cambrian explosion” that we are witnessing in AI chips, the general state of the AI semiconductor market, and the competitive landscape in deep learning, inference, and software infrastructure in support of AI. Karl has a deep background in HPC and AI, having served in executive roles at Cray, IBM, AMD, and Calxeda, a pioneer of Arm-based system-on-chip (SoC) for servers. Karl is a frequent contributor to Forbes.
In this episode of Industry View, we are delighted to have a rare opportunity to catch up with none other than Pete Ungaro, long time luminary and admired leader in HPC/AI. Mr. Ungaro is a globally recognized technology executive, among the “40 under 40” by Corporate Leader Magazine in 2008, and CEO of the year by Seattle Business Monthly for the year 2006. He was most recently SVP/GM of High Performance Computing (HPC), Mission Critical Systems (MCS), and HPE Labs at HPE. Previously, he was president and CEO of Cray Inc. until its acquisition by HPE. Prior to joining Cray in 2003, Mr. Ungaro served as Vice President of Worldwide Deep Computing Sales for IBM.
In this episode of Industry View, we cover the Cray journey as it became the clear winner in exascale systems, the HPE acquisition, the challenges of delivering a new extreme-scale system during COVID, a look at HPC software, storage, power and cooling, and quantum computing, the opportunities and challenges of AI, and the geopolitics of high tech.
We are starting a new feature, looking at HPC, AI, and other advanced technologies through the lens of industry leaders. In this episode, we have the pleasure of a very lively conversation with Alain Andreoli, a longtime luminary of HPC and IT. Mr. Andreoli was with HPE for more than 7 year where he served as group president and EVP of the Hybrid IT Group, helping shape HPE’s strategy for HPC including the acquisition of SGI in 2017. Earlier he was at Sun Microsystems where he was president of the European operations. He was also a senior executive at Ntt, Oracle, and Texas Instruments.
The Cambrian explosion of AI chips has made it hard to tell what chip is good for what. Venkat Vishwanath, Data Science Team Lead at the Argonne Leadership Computing Facility (ALCF), and a Gordon Bell finalist, joins us to discuss the ALCF AI Testbed. Currently working with systems such as Cerebras, Graphcore, SambaNova, Habana, Groq, Untether, Tenstorrent, Esperanto, and others, the Testbed evaluates accelerators from a usability and performance standpoint.
Cartoon of the week leads to the looming rise of AI and what it means for marketing and how businesses should be responding to the new technology in a way that serves them. Then it’s time to discuss Product, the first of the 4 Ps of marketing.
Tim Crawford, CIO Strategic Advisor and founder of research and advisory firm AVOA, joins us in a discussion of generative AI, data sources, emerging uses of AI in the enterprises, and the complexities of managing and regulating AI.
Adrian Cockcroft, Stephen Perrenod, and Shahin Khan get together in a free-flowing coffee-shop style discussion of future system architecture in supercomputing. The motivation for this episode started during the SC22 conference, where several advances seemed to point to significant changes in system design and optimization. This led to Adrian’s article “SC22: CXL3.0, the Future of HPC Interconnects and Frontier vs. Fugaku” and a deeper dive in his paper: “Supercomputing Predictions: Custom CPUs, CXL3.0, and Petalith Architectures”. Similar threads were discussed over at the @HPCpodcast. At the same time, the well-received and well-discussed paper “Myths and Legends in High-Performance Computing” by Satoshi Matsuoka, Jens Domke, Mohamed Wahib, Aleksandr Drozd, and Torsten Hoefler, instigated a valuable discussion of 12 topics, from major technology areas to specific capabilities in future HPC systems, to application performance. All of that is discussed here flavored with some historical accounts.
In what might become a regular segment, we cover important advances in tech that signal changes in markets and policies. This time, we discuss the iPhone moment in AI and the ensuing AI gold rush, virtual quantum computers, and how silicon photonics can change the chip industry.
The latest news in Quantum Computing, as well as Google’s response to ChatGPT, Bard, IBM cloud’s new AI supercomputer, which also leads to a discussion of IBM.
AI in marketing given all the new advances, marketing as a science, add-on marketing (ice cream cones), and tie-in marketing (Nike-Tiffany campaign) and the risks of “rebranding”.
So many great ideas in tech but how do you assess them scientifically? In “Myths and Legends in High-Performance Computing“, Satoshi Matsuoka, Jens Domke, Mohamed Wahib, Aleksandr Drozd, and Torsten Hoefler tackle 12 important topics, from major technology areas to specific capabilities in future HPC systems, to application performance. They help formulate the right questions, and instigate the important discussions, by posing the topics as myths and legends in an enjoyable and humorous paper. Also check out InsideHPC’s coverage of the article: “Conventional Wisdom Watch: Matsuoka & Co. Take on 12 Myths of HPC.” We caught up with Prof. Matsuoka and Hoefler, one in an airport, to discuss the paper and some of the major topics. Really fun and insightful.
In the first episode of 2023, Shahin and Doug discuss the recent chip announcements and their implications for HPC. Also covered are industry predictions for the year to come that were featured in the InsideHPC article, An AI-Flavored Set of HPC Predictions for 2023, AI for public use, and a promise to invite Prof. Matsuoka to discuss his recent paper on common myths in HPC.
In this year-in-review double-issue episode, we continue what is becoming a tradition, covering some of the notable topics of the past year including: HPC market growth, China, exascale and future of supercomputing, quantum tech, SC22, AI, ACM Turing Award, interconnects, the Nvidia-Arm deal, the Chips and Science Act, HPC software, and fusion energy.
The conversational AI, LaMDA seems to represent a significant advance in AI, bringing up discussions of AI sentience, consciousness, and personhood. It also underscores the urgency of thoughtful social policies based on ethical and legal frameworks. Also discussed is the state of Crypto and NFT: cryptocurrencies and non-fungible tokens. Should we look at them as technologies that might find valid use cases, investment vehicles that require close scrutiny, or both? These are very important topics in our times.
Following his always-anticipated and always-insightful closing keynote at the recent ISC conference, we caught up with Prof. Thomas Sterling to discuss the state of HPC. Dr. Sterling is Professor of Intelligent Systems Engineering at the Indiana University (IU) School of Informatics, Computing, and Engineering, and President and Co-founder of Simultac, a technology company focused on non-von-Neumann memory-based system architectures. Since receiving his Ph.D from MIT as a Hertz Fellowm Dr. Sterling has been a pioneer of parallel processing systems in HPC. His many achievements include the creation in 1994 of the “Beowulf cluster” with Donald Becker at NASA, a system that helped drive the scale-out computing architecture.
Here are the topics and the time-stamp in the podcast when they are discussed:
01:23 Supercomputing “Race”
04:15 HPC in Society
06:40 Climate Change, Controlled Fusion
09:00 HPC’s Role in Informing or Helping Set Social Policy
ISC22, the annual International Supercomputing Conference was held last week in Hamburg, Germany, meeting in person after two years. This is a “postview” of the notable developments at this news-rich event.
Major news since our last (double edition) episode included what’s billed as the fastest AI supercomputer by Google, price hikes on chips by TSMC and Samsung, visualization of a black hole in our own galaxy, and IBM’s ambitious and well-executed quantum computing roadmap. We discuss how an AI supercomputer is different, an unexpected impact of chip shortages and price hikes, what it takes to visualize a black hole, and what IBM’s strategy looks to us from a distance.
Jack Dongarra is a leader in supercomputing technologies, parallel programming tools and technologies, and linear algebra and numerical algorithms. He holds appointments at the University of Tennessee, Oak Ridge National Laboratory, and the University of Manchester, and is the recipient of several awards and honors.
In a wide ranging discussion, we cover the Turing Award, TOP500, the state of HPC benchmarks, China’s Exascale systems, and future directions in algorithms. We also talk about future of supercomputing and AI systems, reminisce about a period where a proliferation of system architectures provides a fertile ground for experimentation, and discuss whether we are entering a similar era now. This is another episode you’d want to listen to more than once!
This is part 2 of a special 2-episode discussion of AI in Science with Rick Stevens, Associate Laboratory Director and leader of Exascale Computing Initiative at Argonne National Laboratory and Professor at University of Chicago. In addition to the new ways AI can help advance science, we also discuss ethics, bias, robustness, security,and explainability of AI, and whether AI can replace scientists. We end with a snapshot of Quantum Information Science (QIS), a promising area albeit in its earlier stages of development compared to AI.
A special 2-episode discussion of AI in Science with Rick Stevens, Associate Laboratory Director and leader of Exascale Computing Initiative at Argonne National Laboratory and Professor at University of Chicago. Rick also led a series of Town Halls during 2019 focused on the relevance and applications of AI in scientific research. Held at Argonne, Oak Ridge, and Berkeley National Laboratories, the events were attended by over 1,000 scientists and engineers. This is part 1 of our conversation. Join us.
Special guest Richard Stiennon, research analyst and author of Security Yearbook 2021, joins Shahin and Doug to discuss the state of advanced cyberwarfare involving AI and supercomputing, and its potential role in the war in Ukraine.
Web3, IoT/Edge, AI, HPC, Blockchain, Cryptocurrencies, GPUs and Quantum, Cyber Risk, 5G, and BioTech point to opportunities and threats. Why are there so many big technology trends right now? Doug and Shahin discuss a framework to help make sense of why these trends point to important changes, how these trends are related, and what they mean individually and together.
Doug Black and Shahin Khan are joined by Hyperion Research CEO Earl Joseph to discuss Hyperion’s market findings. Topics include traditional HPC, AI, Cloud, the impact of Covid, industry and global perspective, and what to expect in the future.
After SC21, Patrick Kennedy of Serve the Home online publication got quite the scoop, Raja’s Chip Notes Lay Out Intel’s Path to Zettascale, when he met with Raja Koduri, SVP and GM of Intel’s Accelerated Computing Systems and Graphics (AXG) Group, to discuss Zettascale projections and plans, stipulating a 2027 timeframe. Is that realistic when Exascale has not quite been made official? Tune in and let us know what you think.
From COVID to Climate Change, Edge to Exascale, and AI to Autonomy, 2021 was an impactful year for HPC and Supercomputing, leading some of the most notable global technology advances and some of the most exciting business opportunities of our time. This episode is a lightning “year in review 2021” as we look back and look forward. Join us!
Shahin Khan and Doug Black discuss the Metaverse, its consumer and industrial uses, competition for data, avatars and more realistic digital twins, commerce in the metaverse, and why so many parts of it, from immersive graphics and gaming to physics based simulation and 5G, include computationally intensive components. Give it a listen and let us know what you think.
Welcome to the @HPCpodcast where Shahin Khan and Doug Black discuss supercomputing technologies and the applications, markets, and policies that shape them. In this inaugural episode we cover HPC and AI chips and accelerators, startups, and unicorns, topics that are sure to reappear frequently.
The growing impact of technology on policy is why OrionX started looking at global policy issues and rapidly changing technologies make it especially more complex to set policy or write laws that reflect their intent. A case in point is regulating the gig economy, enabled by digitization and new digitally-transformed business models.
Laws are Like Code
Laws are like computer programs. They translate intent into code. But translating the exact intent, nothing more and nothing less, to an exact recipe is very difficult. In software engineers, this called a bug. All programmers grapple with it, quality assurance (QA) processes and debugging tools try to find and fix them, and yet computer glitches are everywhere. Many of them are “known bugs” and many are hidden until the special circumstances that would expose them materialize.
Some laws/codes are well written: clean, clear, well-structured, cover all the unlikely cases. They are not so buggy. I have read my share of legal contracts and, let’s say, too many of them are not so well written at all. They leave too much for interpretation, are silent on possible fateful courses of events, or are they push it too far, unintentionally counterproductive. You’d need a judge to rule what the contract means regardless of what may have been intended. They beg lawsuits. Here is a very funny example of how contracts can be interpreted. And reinterpreted. But laws can hekp or hurt people so real-life cases are not funny.
Technology Makes it More Complicated to Set Policy
Sometimes the code is fine but times change and what the software does is no longer needed. In other occasions, new situations make the code incompatible with how things are done, making it the wrong tool for the task.
Likewise, laws may face situations for which they have no appropriate prescription. This is what new technologies can do, enabling new possibilities that were not fathomed by the law. Rapidly changing technologies make it more complex to set policy or write laws that reflect their intent.
How Do You Regulate the Gig Economy?
If you enjoy this kind of complexity, a case in point is the effort to regulate the gig economy. My perspective on this topic is that those participating in the gig economy are broadly split into two segments: those who actively seek it or those who are forced to pursue it, those who see it as liberating or those who see it as exploitative. We said as much five years ago in our OrionX 2016 Technology Issues and Predictions:
5- The “gig economy” continues to grow as work and labor are better matched
The changing nature of work and traditional jobs received substantial coverage in 2015. The prospect of artificial intelligence that could actually work is causing fears of wholesale elimination of jobs and management layers. On the other hand, employers routinely have difficulty finding talent, and employees routinely have difficulty staying engaged. There is a structural problem here. The “sharing economy” is one approach, albeit legally challenged in the short term. But the freelance and outsourcing approach is alive and well and thriving. In this model, everything is either an activity-sliced project, or a time-sliced process, to be performed by the most suitable internal or external resources. Already, in Silicon Valley, it is common to see people carrying 2-3 business cards as they match their skills and passions to their work and livelihood in a more flexible way than the elusive “permanent” full-time job.
A Glance at The PRO Act
The PRO Act, which was introduced in 2019, has been making its way through the law-making process, and includes the so-called ABC test, first put to experimented in California, to determine if someone performing a task qualifies as an “employee”.
What follows is my understanding of the scene and a couple of links* for further reading.
Together, PRO and ABC aim to:
boost collective bargaining/unionization and
protect those forced into the gig economy by recognizing them as employees if that’s really what they are.
Both are important issues because growing economic inequality threatens stability, and new technology is changing the nature of work. Regardless of what political/social/economic philosophy guides you, they need to be addressed deliberately. Whether you decide to leave them be or take strong action, you must be able to explain why that is the right way.
Unionization is the more clear-cut part. Those in favor and those against tend to have made up their minds for good or bad reasons and can point to historical evidence to support their point of view. Not everyone can explain why this is a good/bad option or why alternatives do/don’t exist but there seem to be just two camps.
The gig economy part makes it all more complex. The Act changes the meaning of employment and the ABC test makes it easier and obligatory to categorize someone as an employee. While being recognized as an employee because you really are one is a good thing, being labeled an employee when you are not can be a very bad thing. In the first case you gain freedoms you deserve. In the second, you lose freedoms you deserve. Livelihoods are impacted in both cases.
So a third camp has emerged: those who are in favor of bigger and stronger unions but are against the ABC test as written. While many of them are passionate about the intent of the Act, they believe its consequences will cause unnecessary harm. They are starting to doubt, wondering that maybe these harmful consequences are in fact intended. Those in favor seem to have none of it, dismissing worries. Those against the whole thing seem to be enjoying what looks like an unforced error by a conflicted adversary.
What are the Right Questions to Ask?
What are the right questions to ask? Well, it is complex to change the meaning of employment when the “future of work” has been a topic for a decade, and when the avalanche that is the post-industrial economy is disrupting all manner of assumptions. So the questions to ask must include:
What does 2050 look like? It’s easy to agree that a lot will have changed.
How do we get there and how can we reduce the impact of such a big transition?
Getting it right will determine national competitiveness for many decades. How can we make sure we get it right?
Key Question: What Does 2050 Look Like?
The transition from now to, say, 2050 is probably the most important responsibility of all governments/societies in the world. Massive changes are ahead driven by technology and climate.It can be an opportunity or a threat, depending on what we do. AI and “bots” are coming in the trillions. They probably won’t change the meaning of “work”, but they will change the meaning of “jobs” and much else, and not to our specification.
We’d better anticipate it and be prepared. The pandemic reminded us how important being prepared is.
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
OrionX is a team of industry analysts, marketing executives, and demand generation experts. With a stellar reputation in Silicon Valley, OrionX is known for its trusted counsel, command of market forces, technical depth, and original content.
From SiliconANGLE theCUBE: Technology analyst Shahin Khan discusses the intersection of HPC with key industry trends such as 5G, IoT, edge, blockchain AI and quantum computing.
In this era of “Too Much Data”, it is imperative for all organizations to understand and have a plan for Edge Computing. We discuss it all with Bill Walker, another old colleague from Sun Microsystems who is Chief Technology Officer for Edge Computing pioneer Tensor Networks .
Edge Computing is important for many reasons, not least because pretty much all the raw data will originate from outside the cloud, and the vast majority of it will be processed outside the cloud.
Edge Computing is also important because it combines all of the trends that we track at OrionX that enable Digital Transformation: IoT as the fountain of data, 5G as the way data gets transmitted, HPC and AI as the way we make sense of data, Blockchain as an important model for transactions, Cryptocurrencies as the way value is assigned and transferred, and accelerator technologies that make it all possible within an economic envelope.
Mark Himelstein, CTO of RISC-V joins us to discuss what we think will be a very important part of the information technology landscape. RISC-V has cracked the code on taking an open source community approach to chip design. We can expect to see it everywhere from embedded IoT, to communication fabrics, storage, AI, HPC, accelerators, and more. Initially created at the University of California at Berkeley, RISC-V has grown over the past 10 years to a position of serious prominence. It is clear that it will remain important in the coming decades.
OrionX is a team of industry analysts, marketing executives, and demand generation experts. With a stellar reputation in Silicon Valley, OrionX is known for its trusted counsel, command of market forces, technical depth, and original content.
Can you cover, in 30 minutes, the technical basics of IoT, 5G, HPC, AI, Blockchain (including cryptocurrencies and smart contracts), and Quantum Computing?
Yes you can! Well, I hope so, anyway, since that is what we did at the HPC-AI Advisory Council Conference – Stanford University edition. These technologies cannot be ignored. They drive the digital infrastructure of the future enterprise. We believe it is a must to know enough all of them, in one place, so they can inform your strategy, corporate and product roadmap, and your narrative.
OrionX works with clients on the impact of Digital Transformation on them, their customers, and their messages. Generally, they want to track, in one place, trends like IoT, 5G, AI, Blockchain, and Quantum Computing. And they want to know what these trends mean, how they affect each other, and when they demand action, and how to formulate and execute an effective plan. This talk is a somewhat technical summary of such discussions. Needless to say, if that describes you, please let us know.
Here’s the Slide Deck
Please take a look at the slides below and let us know what you think, or if you’d like us to take you through them.
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
The OrionX Research team welcomes special guest Dr. Max Henderson of Rigetti Computing to discuss the intersection of AI and Quantum Computing. Topics of discussion include reformulation, feature extraction, linear solvers, optimization, programming languages and development environments, quantum inspired algorithms, underlying technologies for various quantum computers, and short- and mid-term outlook for AI and quantum computing.
OrionX is a team of industry analysts, marketing executives, and demand generation experts. With a stellar reputation in Silicon Valley, OrionX is known for its trusted counsel, command of market forces, technical depth, and original content.
Digital Transformation (DX) is coming and it will impact you and your customers. It is not just another IT upgrade, but a profound shift in what your clients and customers want, and a shift in what you offer them. It is a shift in your business model.
In a new report, OrionX goes deeper to look at the dimension of DX and to characterize this trend. OrionX clients automatically receive the full report. Here is an excerpt.
DX Started with Digitization
The Information Revolution is already changing all aspects of society, blending several technology shifts in the process and creating new ones. Digitization is at the root of it, and as it progresses, its transformative powers will change everything.
Whereas the Industrial Age helped humans to surpass their mechanical strengths, Information Age is helping humans exceed their cognitive strengths. Eventually, all organizations must transform to exploit new opportunities, repel new threats, or simply to operate in a new world.
DX is a Board Room Imperative
DX is a threat unless you turn it into an opportunity. It is the new world order, and a necessary topic for executive level attention, sponsorship, and decision.
DX is Multi-Disciplinary
What makes DX especially challenging is that it must blend several technology shifts at once; not all of them, but the right ones. Technology disruptions caused by IoT, 5G, Cloud, AI, HPC, Blockchain, Cryptocurrencies, Smart Contracts, Cybersecurity, Quantum Computing, Robotics, 3D printing, Virtual and Augmented Reality, and BioTech are all eligible to impact your and your customers.
[…]
The emerging Information Age and its global impact is the umbrella trend that drives the OrionX Research, Market Execution, and Customer Engagement work. Digital Transformation is the key market trend and vehicle towards it.
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
HPC and AI on Wall Street has been a tradition in New York City for a number of years. It’s a unique occasion to discuss the impact and future of financial services and computing. AI is increasingly associated with High Performance Computing because HPC has dealt with massive amount of data for decades and it is also driving Deep Learning (DL), the approach that has revitalized AI. Read our paper on HPC and AI here.
The conference content was excellent, the kind where you don’t want to miss any of the talks/panels. And the sponsoring vendors covered a lot of interesting technologies.
We had a lively conversation on my panel and received excellent feedback from the audience and on social media.
This end-of-conference #blockchain panel is surprisingly well informed and interesting – unusual for this topic!
It’s also a poster child for the value of having a diversity of experience and multiple viewpoints in a conversation.
What you see below is more or less what I posed to the panel. They are meant to get the discussion going and provide different perspectives for the audience. They worked well and I think they are a good top-10 set right now so feel free to use them or redistribute.
Top-10 Topics to Discuss at a Crypto/Blockchain Panel
Why? What’s the big deal?
In two sentences, what’s the big deal with blockchain/crypto?
Is Crypto a revolution or a fad?
For Wall Street, is Crypto an opportunity or a threat?
When do you know you need to use Blockchain?
Blockchain or Crypto?
Can you separate Crypto from Blockchain? Is it still useful if you do?
ICOs
Is Crypto the new way to fund things? Will the SEC and other regulators let it?
Can you separate utility tokens from security tokens or they always kind of blurry?
Political Support
What countries or states do you see buying into the whole thing in a big way?
Libra
What about Libra and CBDCs? Will Libra happen?
How are Central Bank Digital Currencies different?
Is it really a war with central banks that some people think it is?
Apps
What is missing in the Crypto/Blockchain software stack?
Why are the Crypto/Blockchain apps still so hard to use?
Blockchain/Crypto and IoT. Why do you need to combine the two?
Besides Crypto exchanges and wallets, what killer apps do you see?
Security
When do we give up?! When do we go from “best practices” to something more definitive?
Is quantum-safe cryptography something we need to hurry up and do or do we have time? How much time?
Other Coins
Do other coins have a shot outside of permissoned walled gardens?
In which case, do they really need to be on a blockchain?
Digital Assets
What can we expect in digital assets and especially non-fungible tokens (NFTs)?
Smart Contracts
What about smart contracts? Can they go beyond simple instructions and approach the sophistication of program trading or actual legal contracts?
The HPC and AI on Wall Street conference goes to Singapore next in April 2020 before heading back to NYC next fall. We (OrionX) are looking forward to being there.
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
Co-hosts Dan Olds and Shahin Khan review the results of a recent “Artificial Intelligence, Machine Learning, and Deep Learning” survey that OrionX conducted with over 300 high tech, industrial and commercial companies. See the data and find out how these trends may impact organizations’ AI strategies.
OrionX is a team of industry analysts, marketing executives, and demand generation experts. With a stellar reputation in Silicon Valley, OrionX is known for its trusted counsel, command of market forces, technical depth, and original content.
There are more Things than anything else! And they’re going online in droves, adding capability but also vulnerability, and making the Secure Internet of Things (IoT), the super set of the technology trends. Secure IoT is the place with the most interdisciplinary, and therefore, the most difficult challenges. If you believe, as you should, that no-compromise cybersecurity is table stakes, then IoT means “secure IoT” and IoT by itself becomes the ultimate hashtag soup. Here is why.
IoT and Security are End to End
IoT and security require an end-to-end mindset. Today’s solutions are fragmented. Integrating them is hard and requires blending of many technologies and algorithms. It requires a team with a very broad set of expertise who have managed to understand each other and employ each other’s strengths.
#IoT
Things come in many shapes and sizes. We have to make a distinction between Small-t things, sensors and micro devices, Big-T Things, which often form meta things or Systems, or All-caps THINGS, meta systems and the so-called Distributed Autonomous Organizations (DAOs). Managing them in a coherent manner requires an expertise rubber band that has to stretch across from the microscopic to the astronomic:
Size: Tiny Things to large Things
Data: Simple signals to structured data to context-sensitive streams
Mobility: stationary Things to Things that are carried by people to self-propelled Things
Autonomy: Things that are controlled by people to the so-called DAOs
#Cybersecurity
Security is not just a technology and practice but a non-negotiable requirement that cuts across all aspect of tech. Small-t things require a whole a new approach to device and data security all the way up the supply chain, a difficult technological and operational task an area of intense innovation, for example by blending AI, IoT, and Cybersecurity disciplines or pursuing entirely new approaches. Big-T Things remain hard to secure but have more built-in resources and processing capability, making them eligible to be handled like existing IT systems.
Physical #Security
Things, scattered around, are exposed to physical tampering so it is important to physically secure them and detect any unauthorized physical access or near-field or far-field monitoring. As is often said, “security is security” so physical security and cybersecurity must be viewed and implemented as one thing: all under the umbrella of risk and threat management and #sustainability.
#Mobility
Many Things move, either by themselves (like #drones or robots) or carried by other things (pick a form of transportation) or by people (personal device or #wearables). Mobility issues from initial on-boarding to authenticated secure transmission to data streaming and data semantics to over-the-air (OTA) updates all become critical in IoT.
#Blockchain
Distributed Things that generate lots of data, require secure authenticated communications, are possibly subject to compliance regulations, may need digital rights management (#DRM), or need verifiable delivery… are the kind of use case that Blockchain technology can address very well. IoT applications are often natural candidates. (see also the OrionX Download podcast series on Blockchain.)
#BigData
Connected devices generate data. Making sense of that data requires Big Data practices and technologies: data lakes, data science, storage, search, apps, etc.
#AI
When you have so much data, invariably you will find an opportunity for AI (see the OrionX/ai page for a repository of reports), either via #MachineLearning which uses the data, or #ExpertSystems for policy management which decides what to do about the insight that the data generates. AI framewroks, especially the #DeepLearning variety, are #HPC oriented and require knowledge of HPC systems and algorithms.
There will come a time when another emerging trend, #QuantumComputing, will be useful for AI and cybersecurity.
#Robotics
If Things are smart enough and can move, they become robots. Distributed Autonomous Objects (DAOs) use and generate data and require a host of security, processing, control, policy, etc.
#Cloud
All of this processing happens in the cloud somewhere, so all cloud technologies come into play. The required expertise will cover the gamut: app development, streaming data, dev-ops, elasticity, service level assurance, API management, microservices, data storage, multi-cloud reliability, etc.
Legal Framework
IoT provides new kinds of information extraction, manipulation, and control. Just the AI and Robotics components are enough to pose a challenge for existing legal systems. Ultimately, it requires an ethics framework that can be used to create a proper legal system all the while ensuring widespread communication so organizational structures and culture can keep up with it.
Summary
Not every IoT deployment will need all of the above, but as you try to stitch together a strategy for your IoT project, it is imperative to do so with a big picture in mind. IoT is the essence of #DigitalTransformation and touches all aspects of your organization. It’s quite a challenge to make it seamless, and you’ll need specialists, generalists, and not just analysts but also “synthesists”.
In the OrionX 2016 Technology Issues and Predictions blog, we said “If you missed the boat on cloud, you can’t miss it on IoT too”, and “IoT is where Big Data Analytics, Cognitive Computing, and Machine Learning come together for entirely new ways of managing business processes.” Later, in February of 2016, my colleague Stephen Perrenod’s blog IoT: The Ultimate Convergence, further set the scene. Today, we see those predictions have become reality while IoT is poised for even more convergence.
We will cover these topics in future episodes of The OrionX Download podcast, identifying the salient points. As usual, we will go one or two levels below the surface to understand not just what these technologies do, but how they do it and how they came about. That will in turn, help put new developments in perspective and explain why they may or may not address a pain point.
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
We are very pleased to announce the completion of our “epic” AI survey. Here is the press release that we issued this morning:
ORIONX RESEARCH ANNOUNCES SURVEY RESULTS REVEALING CUSTOMERS’ ADOPTION OF ARTIFICIAL INTELLIGENCE (AI), MACHINE LEARNING (ML) AND DEEP LEARNING (DL)
MENLO PARK, Calif., August 30, 2017 – OrionX Research today announced the availability of survey results of more than 300 North American companies representing 13 industries on what they are currently doing and planning to do with Artificial Intelligence (AI), Machine Learning (ML) and Deep Learning (DL) technologies.
Considered one of the most comprehensive surveys to date on AI/ML/DL, the survey covers 144 questions/data points, presented in over 70 charts and graphs. It explores topics ranging from key drivers for AI project implementation to customer buying behavior to current and future AI budget spend to the most popular AI models.
Study highlights include:
Top challenges businesses are facing for their AI/ML/DL projects
Current and planned budgets for AI/ML/DL projects (broken down by hardware, software, services, cloud, consulting, education/training, staffing and more)
Who the decision makers are and which departments are driving AI purchasing decisions
Average AI project data size (TB) and the application areas that AI projects are being used for
User ranking of over 20 AI Frameworks with TensorFlow dominating at 52%
AI hardware usage (now and in the future), including most used hardware accelerators with NVIDIA in the lead at 82%
Customers’ key selection criteria and requirements when evaluating AI vendors
“The OrionX AI survey is the first to assess customer sentiment across a wide range of AI adoption issues,” said Shahin Khan, founding partner at OrionX. “Among many insights, it shows how customer selection criteria is influenced by parameters such as vendor market presence, product roadmap, current product features, ease of adoption, ease of use, interoperability, security, performance, and robustness.” This comprehensive AI study will help customers better understand the AI/ML/DL landscape in order to set strategy and guide future investments.
To purchase the report and learn how OrionX research analysts can help companies navigate the multi-faceted and rapidly evolving AI/ML/DL marketplace, visit www.orionx.net/ai
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
Here at OrionX.net, our research agenda is driven by the latest developments in technology. We are also fortunate to work with many tech leaders in nearly every part of the “stack”, from chips to apps, who are driving the development of such technologies.
In recent years, we have had projects in Cryptocurrencies, IoT, AI, Cybersecurity, HPC, Cloud, Data Center, … and often a combination of them. Doing this for several years has given us a unique perspective. Our clients see value in our work since it helps them connect the dots better, impacts their investment decisions and the options they consider, and assists them in communicating their vision and strategies more effectively.
We wanted to bring that unique perspective and insight directly to you. “Simplifying the big ideas in technology” is how we’d like to think of it.
The OrionX Download is both a video slidecast (visuals but no talking heads) and an audio podcast in case you’d like to listen to it.
Every two weeks, co-hosts Dan Olds and Shahin Khan, and other OrionX analysts, discuss some of the latest and most important advances in technology. If our research has a specially interesting finding, we’ll invite guests to probe the subject and add their take.
Please give it a try and let us know what we can do better or if you have a specific question or topic that you’d like us to cover.
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
Just as Intel, the king of CPUs and the very bloodstream of computing announced that it is ending its Intel Developer Forum (IDF) annual event, this week in San Jose, NVIDIA, the king of GPUs and the fuel of Artificial Intelligence is holding its biggest GPU Technology Conference (GTC) annual event yet. Coincidence? Hardly.
With something north of 95 per cent market share in laptops, desktops, and servers, Intel-the-company is far from even looking weak. Indeed, it is systematically adding to its strengths with a strong x86 roadmap, indigenous GPUs of its own, acquisition of budding AI chip vendors, pushing on storage-class memory, and advanced interconnects.
But a revolution is nevertheless afoot. The end of CPU-dominated computing is upon us. Get ready to turn the page.
The Rise of AI
Digitization means lots of data, and making sense of lots of data increasingly looks like either an AI problem or an HPC problem (which are similar in many ways, see “Is AI the Future of HPC?”). Either way, it includes what we call High Density Processing: GPUs, FPGAs, vector processors, network processing, or other, new types of accelerators.
GPUs made deep neural networks practical, and it turns out that after decades of slow progress in AI, such Deep Learning algorithms were the missing ingredient to make AI effective across a range of problems.
NVIDIA was there to greet this turn of events and has ridden it to strong leadership of a critical new wave. It’s done all the right things and executed well in its usual competent manner.
Competition
A fast-growing lucrative market means competition, of course, and a raft of new AI chips is around the corner.
Intel already acquired Nervana and Movidius. Google has its TPU, and IBM its neuromorphic chip, TrueNorth. Other AI chip efforts include Mobileye (the company is being bought by Intel), Graphcore, BrainChip, TeraDeep, KnuEdge, Wave Computing, and Horizon Robotics. In addition, there are several well-publicized and respected projects like NeuRAM3, P-Neuro, SpiNNaker, Eyeriss, and krtkl going after different parts of the market.
One thing that distinguishes Nvidia is how it addresses several markets with pretty much a single underlying platform. From automobiles to laptops to desktop to gaming to HPC to AI, the company manages to increase its Total Available Market (TAM) with minimal duplication of effort. It remains to be seen whether competitive chips that are super optimized for just one market can get enough traction to pose a serious threat.
The world of computing is changing in profound ways and chips are once again an area of intense innovation. The Silicon in Silicon Valley is re-asserting itself in a most welcome manner.
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
The data center market is hot, especially now that we are getting a raft of new stuff, from promising non-Intel chips and system architectures to power and cooling optimizations to new applications in Analytics, IoT, and Artificial Intelligence.
Here is my Top-10 Data Center Predictions for 2017:
1) Data center optimization is here
Data centers are information factories with lots of components and moving parts. There was a time when companies started becoming much more complex, which fueled the massive enterprise resource planning market. Managing everything in the data center is in a similar place. To automate, monitor, troubleshoot, plan, optimize, cost-contain, report, etc, is a giant task, and it is good to see new apps in this area.
Data center infrastructure management provides visibility into and helps control IT. Once it is deployed, you’ll wonder how you did without it. Some day, it will be one cohesive thing, but for now, because it’s such a big task, there will be several companies addressing different parts of it.
2) Azure will grow faster than AWS
Cloud is the big wave, of course, and almost anything that touches it is on the right side of history. So, private and hybrid clouds will grow nicely and they will even temper the growth of public clouds. But the growth of public clouds will continue to impress, despite increasing recognition that they are not the cheapest option, especially for the mid-size users.
AWS will lead again, capturing most new apps. However, Azure will grow faster, on the strength of both landing new apps and also bringing along existing apps where Microsoft maintains a significant footprint.
Moving Exchange, Office, and other apps to the cloud, in addition to operating lots of regional data centers and having lots of local feet on the ground will help.
Some of the same dynamics will help Oracle show a strong hand and get close to Google and IBM. And large telcos will stay very much in the game. Smaller players will persevere and even grow, but they will also start to realize that public clouds are their supplier or partner, not their competition! It will be cheaper for them to OEM services from bigger players, or offer joint services, than to build and maintain their own public cloud.
3) Great Wall of persistent memory will become a thing
Just as the hottest thing in enterprise becomes Big Data, we find that the most expensive part of computing is moving all that data around. Naturally, we start seeing what OrionX calls “In-Situ Processing” (see page 4 of the report in the link): instead of data going to compute, compute would go to data, processing it locally wherever data happens to be.
But as the gap between CPU speed and storage speed separates apps and data, memory becomes the bottleneck. In comes storage class memory (mostly flash, with a nod to other promising technologies), getting larger, faster and cheaper. So, we will see examples of apps using a Great Wall of persistent memory, built by hardware and software solutions that bridge the size/speed/cost gap between traditional storage and DRAM. Eventually, we expect programming languages to naturally support byte-addressable persistent memory.
4) System vendors will announce racks, not servers
Vendors already configure and sell racks, but they often populate them with servers that are designed as if they’d be used stand-alone. Vendors with rack-level thinking have been doing better because designing the rack vs the single node lets them add value to the rack while removing unneeded value from server nodes.
So server vendors will start thinking of a rack, not a single-node, as the system they design and sell. Intel’s Rack Scale Architecture has been on the right track, a real competitive advantage, and an indication of how traditional server vendors must adapt. The server rack will become the next level of integration and what a “typical system” will look like. Going forward, multi-rack systems are where server vendors have a shot at adding real value. HPC vendors have long been there.
5) Server revenue growth will be lower than GDP growth
Traditional enterprise apps – the bulk of what runs on servers – will show that they have access to enough compute capacity already. Most of that work is transactional, so their growth is correlated with the growth in GDP, minus efficiencies in processing.
New apps, on the other hand, are hungry, but they are much more distributed, more focused on mobile clients, and more amenable to what OrionX calls “High-Density Processing”: algorithms that have a high ops/bytes ratio running on hardware that provides similarly high ops/byte capability – ie, compute accelerators like GPUs, FPGAs, vector processors, manycore CPUs, ASICs, and new chips on the horizon.
On top of that, there will be more In-Situ Processing: processing the data wherever it happens to be, say locally on the client vs sending it around to the backend. This will be made easier by the significant rise in client-side computing power and more capable data center switches and storage nodes that can do a lot of local processing.
We will also continue to see cloud computing and virtualization eliminate idle servers and increase the utilization rates of existing systems.
Finally, commoditization of servers and racks, driven by fewer-but-larger buyers and standardization efforts like the Open Compute Project, put pressure on server costs and limit the areas in which server vendors can add value. The old adage in servers: “I know how to build it so it costs $1m, but don’t know how to build it so it’s worth $1m” will be true more than ever.
These will all combine to keep server revenues in check. We will see 5G’s wow-speeds but modest roll-out, and though it can drive a jump in video and some server-heavy apps, we’ll have to wait a bit longer.
6) OpenPOWER will emerge as the viable alternative to x86
The real battle in server architecture will be between Intel’s in-house coalition and what my colleague Dan Olds has called the Rebel Alliance: IBM’s OpenPower industry coalition. Intel brings its all-star team: Xeon Phi, Altera, Omni-Path (plus Nervana/Movidius) and Lustre, while OpenPower counters with a dream team if its own: POWER, Nvidia, Xilinx, and Mellanox (plus TrueNorth) and GPFS (Spectrum Scale). Then there is Watson which will become more visible as a differentiator, and a series of acquisitions by both camps as they fill various gaps.
The all-in-house model promises seamless integration and consistent design, while the extended team offers a best-of-breed approach. Both have merits. Both camps are pretty formidable. And there is real differentiation in strategy, design, and implementation. Competition is good.
7) Beware IoT Security
One day in the not too distant future, your fancy car will break down on the highway for no apparent reason. It will turn out that the auto entertainment system launched a denial of service attack on the rest of the car, in a bold attempt to gain control. It even convinced the nav system to throw in with it! This is a joke, of course, but could become all too real if/when human and AI hackers get involved.
IoT is coming, and with it will come all sorts of security issues. Do you know where your connected devices are? Can you really trust them?
8) More chips than Vegas; riskier too
For the first time in decades, there is a real market opening for new chips. What drove this includes:
The emergence of new apps, led by AI and IoT. The part of AI that is computationally interesting and different is High-Performance AI, since it intersects with HPC. On the IoT side, backend apps are typically Analytics to make sense of sensor data. These new apps will run where they can run better/faster/cheaper. They will be too new to be burdened by any allegiance or bonds to a particular chip.
The fact that many existing apps have no clue what hardware they run on, and operate on the upper layers of a tall stack.
The possibility to build a complete software stack from open source software components.
The presence of very large customers like cloud providers or top supercomputing sites. They buy in seriously large volumes and have the wherewithal to build the necessary software stack, so they can afford to pick a new chip and bolster, if not guarantee, its viability.
This will be a year when many new chips became available and tested, and there is a pretty long list of them, showing just how big the opportunity is, how eager investors must have been to not miss out, and how many different angles there are.
In addition to AI, there are a few important general-purpose chips being built. The coolest one is by startup Rex Computing, which is working on a chip for exascale, focused on very high compute-power/electric-power ratios. Qualcomm and Caviuum have manycore ARM processors, Oracle is advancing SPARC quite well, IBM’s POWER continues to look very interesting, and Intel and AMD push the X86 envelope nicely.
With AI chips, Intel already has acquired Nervana and Movidius. Google has its TPU, and IBM its neuromorphic chip, TrueNorth. Other AI chip efforts include Mobileye (the company is being bought by Intel since this article was written), Graphcore, BrainChip, TeraDeep, KnuEdge, Wave Computing, and Horizon Robotics. In addition, there are several well-publicized and respected projects like NeuRAM3, P-Neuro, SpiNNaker, Eyeriss, and krtkl going after different parts of the market. That’s a lot of chips, but most of these bets, of course, won’t pay off.
9) ARM server market share will stay below 3 per cent
Speaking of chips, ARM servers will remain important but elusive. They will continue to make a lot of noise and point to significant wins and systems, but fail to move the needle when it comes to revenue market share in 2017.
As a long-term play, ARM is an important phenomenon in the server market – more so now with the backing of SoftBank, a much larger company apparently intent on investing and building, and various supercomputing and cloud projects that are great proving grounds.
But at the end, you need differentiation and ARM has the same problem against X86 as Intel’s Atom had against ARM. Atom did not differentiate vs ARM any more than ARM is differentiating vs Xeon.
Most systems end up being really good at something, however, and there are new apps and an open-source stack to support existing apps, which will help find specific workloads where ARM implementations could shine. That will help the differentiation become more than “it’s not X86”.
10) Is it an app, or is it a fabric? More cloud fabrics introduced
What is going on with big new apps? They keep getting more modular (microservices), more elastic (scale out), and more real-time (streaming). They’ve become their own large graph, in the computer science sense, and even more so with IoT (sensors everywhere plus In-Situ Processing).
As apps became services, they began resembling a network of interacting modules, depending on each other and evolving together. When the number of interacting pieces keeps increasing, you’ve got yourself a fabric. But as an app, it’s the kind of fabric that evolves and has an overall purpose (semantic web).
Among engineering disciplines, software engineering already doesn’t have a great reputation for predictability and maintainability. More modularity is not going to help with that.
But efforts to manage interdependence and application evolution have already created standards for structured modularity like the OSGi Alliance for Java. Smart organizations have had ways to reduce future problems (technical debt) from the get-go. So, it will be nice to see that type of methodology get better recognized and better generalized.
OK, we stop here now, leaving several interesting topics for later.
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
What does a Spanish silver dollar have to do with Deep Learning? It’s a question of standards and required precision.The widely used Spanish coin was introduced at the end of the 16th century as Spain exploited the vast riches of New World silver. It was denominated as 8 Reales. Because of its standard characteristics it served as a global currency.
Ferdinand VI Silver peso (8 Reales, or Spanish silver dollar)
The American colonies in the 18th century suffered from a shortage of British coinage and used the Spanish dollar widely; it entered circulation through trade with the West Indies. The Spanish dollar was also known as “pieces of eight” and in fact was often cut into pieces known as “bits” with 8 bits comprising a dollar. This is where the expression “two bits” referring to a quarter dollar comes from. The original US dollar coin was essentially based on the Spanish dollar.
For Deep Learning, the question arises – what is the requisite precision for robust performance of a multilayer neural network. Most neural net applications are implemented with 32 bit floating point precision, but is this really necessary?
It seems that many neural net applications could be successfully deployed with integer or fixed point arithmetic rather than floating point, and with only 8 to 16 bits of precision. Training may require higher precision, but not necessarily.
A team of researchers from IBM’s Watson Labs and Almaden Research Center find that:
“deep networks can be trained using only 16-bit wide fixed-point number representation when using stochastic rounding, and incur little to no degradation in the classification accuracy”.
“As long as you accumulate to 32 bits when you’re doing the long dot products that are the heart of the fully-connected and convolution operations (and that take up the vast majority of the time) you don’t need float though, you can keep all your inputs and output as eight bit. I’ve even seen evidence that you can drop a bit or two below eight without too much loss! The pooling layers are fine at eight bits too, I’ve generally seen the bias addition and activation functions (other than the trivial relu) done at higher precision, but 16 bits seems fine even for those.”
He goes on to say that “training can also be done at low precision. Knowing that you’re aiming at a lower-precision deployment can make life easier too, even if you train in float, since you can do things like place limits on the ranges of the activation layers.”
Moussa and co-researchers have found 12 times greater speed using a fixed-point representation when compared to floating point on the same Xilinx FPGA hardware. If one can relax the precision of neural nets when deployed and/or during training, then higher performance may be realizable at lower cost and with a lower memory footprint and lower power consumption. The use of heterogeneous architectures employing GPUs, FPGAs or other special purpose hardware becomes even more feasible.
This is such an interesting area, with manycore chips such as Intel’s Xeon Phi, nVidia’s GPUs and various FPGAs jockeying for position in the very hot Deep Learning marketplace.
OrionX will continue to monitor AI and Deep Learning developments closely.
References:
Gupta, S., Agrawal, A., Gopalakrishnan, K., Narayanan, P. 2015,https://arxiv.org/pdf/1502.02551.pdf “Deep Learning with Limited Numerical Precision”
Jin, L. et al. 2014, “Training Large Scale Deep Neural Networks on the Intel Xeon Phi Many-Core Processor”, proceedings, Parallel & Distributed Processing Symposium Workshops, May 2014
Moussa, M., Areibi, S., and Nichols, K. 2006 “Arithmetic Precision for Implementing BP Networks on FPGA: A case study”, Chapter 2 in FPGA Implementations of Neural Networks, ed. Omondi, A. and Rajapakse, J.
Stephen Perrenod has lived and worked in Asia, the US, and Europe and possesses business experience across all major geographies in the Asia-Pacific region. He specializes in corporate strategy for market expansion, and cryptocurrency/blockchain on a deep foundation of high performance computing (HPC), cloud computing and big data. He is a prolific blogger and author of a book on cosmology.
On the Monday of the conference, a new leader on the TOP500 list was announced. The Sunway TaihuLight system uses a new processor architecture that is Single-Instruction-Multiple-Data (SIMD) with a pipeline that can do eight 64-bit floating-point calculations per cycle.
This started us thinking about vector processing, a time-honored system architecture that started the supercomputing market. When microprocessors advanced enough to enable massively parallel processing (MPP) systems and then Beowulf and scale-out clusters, the supercomputing industry moved away from vector processing and led the scale-out model.
Later that day, at the “ISC 2016 Vendor Showdown”, NEC had a presentation about its project “Aurora”. This project aims to combine x86 clusters and NEC’s vector processors in the same high bandwidth system. NEC has a long history of advanced vector processors with its SX architecture. Among many achievements, it built the Earth Simulator, a vector-parallel system that was #1 on the TOP500 list from 2002 to 2004. At its debut, it had a substantial (nearly 5x) lead over the previous #1.
Close integration of accelerator technologies with the main CPU is, of course, a very desirable objective. It improves programmability and efficiency. Along those lines, we should also mention the Convey system, which goes all the way, extending the X86 instruction set, and performing the computationally intensive tasks in an integrated FPGA.
A big advantage of vector processing is that it is part of the CPU with full access to the memory hierarchy. In addition, compilers can do a good job of producing optimized code. For many codes, such as in climate modelling, vector processing is quite the right architecture.
Vector parallel systems extended the capability of vector processing and reigned supreme for many years, for very good reasons. But MPPs pushed vector processing back, and GP-GPUs pushed it further still. GPUs leverage the high volumes that the graphics market provides and can provide numerical acceleration with some incremental engineering.
But as usual, when you scale more and more, you scale not just capability, but also complexity! Little inefficiencies start adding up until they become a serious issue. At some point, you need to revisit the system and take steps, perhaps drastic steps. The Sunway TaihuLight system atop the TOP500 list is an example of this concept. And there are new applications like deep learning that look like they could use vectors to quite significant advantage.
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
The world of quantum computing frequently seems as bizarre as the alternate realities created in Lewis Carroll’s masterpieces “Alice’s Adventures in Wonderland” and “Through the Looking-Glass”. Carroll (Charles Lutwidge Dodgson) was a well-respected mathematician and logician in addition to being a photographer and enigmatic author.
Has quantum computing’s time actually come or are we just chasing rabbits?
That is probably a twenty million dollar question by the time a D-Wave 2X™ System has been installed and is in use by a team of researchers. Publicly disclosed installations currently include Lockheed Martin, NASA’s Ames Research Center and Los Alamos National Laboratory.
Hosted at NASA’s Ames Research Center in California, the Quantum Artificial Intelligence Laboratory (QuAIL) supports a collaborative effort among NASA, Google and the Universities Space Research Association (USRA) to explore the potential for quantum computers to tackle optimization problems that are difficult or impossible for traditional supercomputers to handle. Researchers on NASA’s QuAIL team are using the system to investigate areas where quantum algorithms might someday dramatically improve the agency’s ability to solve difficult optimization problems in aeronautics, Earth and space sciences, and space exploration. For Google the goal is to study how quantum computing might advance machine learning. The USRA manages access for researchers from around the world to share time on the system.
Using quantum annealing to solve optimization problems
D-Wave’s quantum annealing technology addresses optimization and probabilistic sampling problems by framing them as energy minimization problems and exploiting the properties of quantum physics to identify the most likely outcomes or as a probabilistic map of the solution landscape.
Quantum annealer dynamics are dominated by paths through the mean field energy landscape that have the highest transition probabilities. Figure 1 shows a path that connects local minimum A to local minimum D.
Figure 2 shows the effect of quantum tunneling (in blue) to reduce the thermal activation energy needed to overcome the barriers between the local minima with the greatest advantage observed from A to B and B to C, and a negligible gain from C to D. The principle and benefits are explained in detail in the paper “What is the Computational Value of Finite Range Tunneling?”
The D-Wave 2X: Interstellar Overdrive – How cool is that?
As a research area, quantum computing is highly competitive, but if you want to buy a quantum computer then D-Wave Systems , founded in 1999, is the only game in town. Quantum computing is as promising as it is unproven. Quantum computing goes beyond Moore’s law since every quantum bit (qubit) doubles the computational power, similar to the famous wheat and chessboard problem. So the payoff is huge, even though it is expensive, unproven, and difficult to program.
The advantage of quantum annealing machines is they are much simpler to build than gate-model quantum computing machines. The latest D-Wave machine (the D-Wave 2X), installed at NASA Ames, is approximately twice as powerful (in a quantum, exponential sense) as the previous model at over 1,000 qubits (1,097). This compares with roughly 10 qubits for current gate-model quantum systems, so two orders of magnitude. It’s a question of scale, no simple task, and a unique achievement. Although quantum researchers initially questioned whether the D-Wave system even qualified as a quantum computer, albeit a subset of quantum computing architectures, that argument seems mostly settled and it is now generally accepted that quantum characteristics have been adequately demonstrated.
In a D-Wave system, a coupled pair of qubits (quantum bits) demonstrate quantum entanglement (they influence each other), so that the entangled pair can be in any one of four states resulting from how the coupling and energy biases are programmed. By representing the problem to be addressed as an energy map the most likely outcomes can be derived by identifying the lowest energy states.
A lattice of approximately 1,000 tiny superconducting circuits (the qubits) is chilled close to absolute zero to deliver quantum effects. A user models a problem into a search for the lowest point in a vast energy landscape. The processor considers all possibilities simultaneously to determine the lowest energy required to form those relationships. Multiple solutions are returned to the user, scaled to show optimal answers, in an execution time of around 20 microseconds, practically instantaneously for all intents and purposes.
The D-wave system cabinet – “The Fridge”– is a closed cycle dilution refrigerator. The superconducting processor itself generates no heat, but to operate reliably must be cooled to about 180 times colder than interstellar space, approximately 0.015° Kelvin.
Environmental considerations: Green is the color
To function reliably, quantum computing systems require environments that are not only shielded from the Earth’s natural environment, but would be considered inhospitable to any known life form. A high vacuum is required, a pressure 10 billion times lower than atmospheric pressure, and shielded to 50,000 times less than Earth’s magnetic field. Not exactly a normal office, datacenter, or HPC facility environment.
On the other hand, the self-contained “Fridge” and servers consume just 25kW of power (approximately the power draw of a single heavily populated standard rack) and about three orders of magnitude (1000 times) less power than the current number one system on the TOP500, including its cooling system. Perhaps a more significant consideration is that power demand is not anticipated to increase significantly as it scales to several thousands of qubits and beyond.
In addition to doubling the number of qubits compared with the prior D-Wave system, the D-Wave 2X delivers lower noise in qubits and couples, delivering greater confidence in achieved results.
So much for the pictures, what about the conversations?
Now that we have largely moved beyond the debate of whether a D-Wave system is actually a quantum machine or not, then the question “OK, so what now?” could bring us back to chasing rabbits, although this time inspired by the classic Jefferson Airplane song, “White Rabbit”:
“One algorithm makes you larger
And another makes you small
But do the ones a D-Wave processes
Do anything at all?”
That of course, is where the conversations begin. It may depend upon the definition of “useful” and also a comparison between “conventional” systems and quantum computing approaches. Even the fastest supercomputer we can build using the most advanced traditional technologies can still only perform analysis by examining each possible solution serially, one solution at a time. This makes optimizing complex problems with a large number of variables and large data sets a very time consuming business. By comparison, once a problem has been suitably constructed for a quantum computer it can explore all the possible solutions at once and instantly identify the most likely outcomes.
If we consider relative performance then we begin to have a simplistic basis for comparison, at least for execution times. The QuAIL system was benchmarked for the time required to find the optimal solution with 99% probability for different problem sizes up to 945 variables. Simulated Annealing (SA), Quantum Monte Carlo (QMC) and the D-Wave 2X were compared. Full details are available in the paper referenced previously. Shown in the chart are the 50th, 75th and 85th percentiles over a set of 100 instances. The error bars represent 95% confidence intervals from bootstrapping.
This experiment occupied millions of processor cores for several days to tune and run the classical algorithms for these benchmarks. The runtimes for the higher quantiles for the larger problem sizes for QMC were not computed because the computational cost was too high.
The results demonstrate a performance advantage to the quantum annealing approach by a factor of 100 million compared with simulated annealing running on a single state of the art processor core. By comparison the current leading system on the TOP500 has fewer than 6 million cores of any kind, implying a clear performance advantage for quantum annealing based on execution time.
The challenge and the next step is to explore the mapping of real world problems to quantum machines and to improve the programming environments, which will no doubt take a significant amount of work and many conversations. New players will become more visible, early use cases and gaps will become better defined, new use cases will be identified, and a short stack will emerge to ease programming. This is reminiscent of the early days of computing or space flight.
A quantum of solace for the TOP500: Size still matters.
Even though we don’t expect to see viable exascale systems this decade, and quite likely not before the middle of the next, we won’t be seeing a Quantum500 anytime soon either. NASA talks about putting humans on Mars sometime in the 2030s and it isn’t unrealistic to think about practical quantum computing as being on a similar trajectory. Recent research at the University of New South Wales (UNSW) in Sidney, Australia demonstrated that it may be possible to create quantum computer chips that could store thousands, even millions of qubits on a single silicon processor chip leveraging conventional computer technology.
Although the current D-Wave 2X system is a singular achievement it is still regarded as being relatively small to handle real world problems, and would benefit from getting beyond pairwise connectivity, but that isn’t really the point. It plays a significant role in research into areas such as vision systems, artificial intelligence and machine learning alongside its optimization capabilities.
In the near term, we’ve got enough information and evidence to get the picture. It will be the conversations that become paramount with both conventional and quantum computing systems working together to develop better algorithms and expand the boundaries of knowledge and achievement.
An important part of the OrionX magic is the ability to understand technologies, the competitive landscape, and customer needs. We put a lot of effort into that. That is what clients need to connect products to customers, and to meet real customer needs with genuine product capabilities.
Today we raise the bar in that effort as we announce Dan’s addition to the OrionX team and celebrate a new milestone. In addition to tracking key technologies and emerging market segments, Dan has built a well-established process to measure customer sentiment in enterprise IT, something every technology vendor needs.
Thanks to the support of all of you: our clients, colleagues, and community, OrionX is poised to grow its reach and expand its services in Strategy, Marketing, and PR. We’ve been lucky to know you and to work on leading edge technologies in important areas such as IoT, Cloud, Big Data, HPC, & Security, etc.
Here is the press release we issued today:
ORIONX APPOINTS BIG DATA & HPC EXPERT DAN OLDS AS PARTNER
March 2nd, 2016, Menlo Park, CA – Pioneering a new consulting services model for strategy, marketing, and PR, OrionX today announced the appointment of Dan Olds as partner. Mr. Olds is principal analyst at Gabriel Consulting Group (GCG) a Portland, Oregon-based IT consulting firm, which he founded in 2002 and whose activities continue as part of OrionX.
“Our clients value OrionX’s ability to understand their technology, their competition, and their customers, as they pave the way for digital transformation,” said Shahin Khan, Founding Partner at OrionX. “We are raising the bar with Dan’s extensive knowledge of technology trends and insightful appreciation of customer adoption patterns.”
An authority on technology trends and customer requirements, Dan Olds is a frequently quoted expert in industry and business publications such as The Wall Street Journal, Bloomberg News, Computerworld, eWeek, CIO.com, and PCWorld.
“I am thrilled to join OrionX and a group of professionals with whom I have collaborated closely in the past,” said Dan Olds, Partner at OrionX. “My work on technology trends and customer sentiment is a perfect match with OrionX as we expand our services to deliver sound advice and quality content to our clients.”
In addition to server, storage, and network technologies, Dan closely follows the Big Data, Cloud, and HPC markets. He writes the “HPC Blog” on The Register, co-hosts the popular Radio Free HPC podcast, and is the acknowledged go-to person for the coverage of the supercomputing industry’s highly lauded Student Cluster Challenge.
Big Data and HPC Expert Dan Olds Joins OrionX
About OrionX
Mobile, Social, Cloud, Big Data and IoT are redefining B2B technology. OrionX.net is the pioneer of a new model for Strategy, Marketing, and PR. OrionX is known for its depth and agility, insightful advice, original content, and its stellar reputation in Silicon Valley. More than 40 technology leaders in virtually every B2B segment have trusted OrionX.net to help set new break-away strategies, ignite brands, and engage customers for increased market share. Visit us at OrionX.net
For more information: Cindee Mock, OrionX, cindee.mock@orionx.net 650-255-2975
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
In our 2016 predictions blog, we said “If you missed the boat on cloud, you can’t miss it on IoT too”, and “IoT is where Big Data Analytics, Cognitive Computing, and Machine Learning come together for entirely new ways of managing business processes.”
IoT represents the ultimate convergence theme in the marketplace today. IoT includes fixed and mobile edge devices (fog computing), cloud computing, big data, analytics and machine learning, and in many instances can include social and mobile computing as well.
Composition VIII, Wassily Kandinsky, 1923, Shared via #WikiArtApp http://bit.ly/18GCSaj
Because of the very large number of devices being incorporated in IoT solutions and because of the high data rates that must be supported at the very edge of a network, IoT also requires embedded computing with low-power processors for preliminary data ingestion and filtering. Data processing at the edge accomplishes two things: firstly, it allows elimination of data that does not need to be transmitted to the central cloud-based data lake. Secondly, it supports preliminary real-time processing of acquired data that can be used for device monitoring and control with immediate feedback and very low latency. Additionally, computing at the edge should be a prerequisite for enabling security of devices and the data at the time of acquisition.
IOT and cloud computing fit together like a hand in a glove. This is only natural with the highly distributed and very dynamic nature of IoT devices and IoT data. Cloud infrastructure provides the most versatile, flexible and adaptable platform for an IoT repository and processing system. Cloud vendors including Amazon (AWS), Microsoft (Azure), IBM (SoftLayer) and Alphabet (Google Compute Engine) are racing to implement IoT capabilities, APIs, and advanced analytics and machine learning solutions in their public cloud environments. For reasons of control and to meet privacy and regulatory requirements, many companies will choose to implement private cloud services as well. Hybrid cloud services will play an important role in IoT.
When it comes to infrastructure within IoT clouds and at the edge, one can imagine every type of processor, network and protocol, servers and storage playing a role – both in distributed and centralized fashions – for IoT environments. Shared APIs and protocols are critical to promote interoperability. End-to-end security is as well an imperative; security is needed within every device and at each and every layer and sub-layer of the solution.
If anything deserves the term Big Data it is IoT. Billions and billions of devices are becoming Internet-enabled and the number of devices engaged in IoT applications will soon exceed the number of people on Earth. Because there are now many Big Data solutions available in public clouds this reinforces the cloud as a natural repository for IoT-generated Big Data lakes.
Analytics and machine learning will be heavily used to extract maximum value from the large amount of data acquired. A very wide range of business intelligence, analytics and machine learning techniques will be required. We foresee that IoT will be a big driver for machine learning advances. Analytics and machine learning will support everything from better operations of the connected devices to better information on how devices are utilized and new value-added services enabled by the device manufacturers.
There is a wide array of potential IoT applications covering every industry: Transportation, Retail, Manufacturing, Energy, Finance, Smart Cities, Healthcare, Agriculture, Government and other areas. Consumer apps are found in areas such as wearables, fitness, smart homes, electronics, and gaming.
Benefits from IoT applications will include improved operational efficiency and reliability, insight into usage patterns, and opportunities for increased revenue and better product design for manufacturers. Just as today we cannot imagine life without the Internet, we will not be able to imagine life without IoT by the beginning of the next decade.
It’s difficult to think of an area that requires a more holistic and converged view than IoT and that is not already a natural subset or example of IoT. Robotics? Drones? 3D printing? …. Every device of value should benefit from being plugged into the Internet, if only for maintenance and monitoring purposes, and these IoT-enabled devices will become available to participate in other IoT apps. At the other end of the data flow, big value comes from analytics services that are applied to a pool of devices and the streaming data they provide.
IT and IoT will become inseparable. IoT requires, incorporates, and benefits from all of the other major themes in IT today and thus we see it as representing the Ultimate Convergence at present.
Stephen Perrenod has lived and worked in Asia, the US, and Europe and possesses business experience across all major geographies in the Asia-Pacific region. He specializes in corporate strategy for market expansion, and cryptocurrency/blockchain on a deep foundation of high performance computing (HPC), cloud computing and big data. He is a prolific blogger and author of a book on cosmology.
Here at OrionX.net, we are fortunate to work with tech leaders across several industries and geographies, serving markets in Mobile, Social, Cloud, and Big Data (including Analytics, Cognitive Computing, IoT, Machine Learning, Semantic Web, etc.), and focused on pretty much every part of the “stack”, from chips to apps and everything in between. Doing this for several years has given us a privileged perspective. We spent some time to discuss what we are seeing and to capture some of the trends in this blog: our 2016 technology issues and predictions. We cut it at 17 but we hope it’s a quick read that you find worthwhile. Let us know if you’d like to discuss any of the items or the companies that are driving them.
1- Energy technology, risk management, and climate change refashion the world
Energy is arguably the most important industry on the planet. Advances in energy efficiency and sustainable energy sources, combined with the debate and observations of climate change, and new ways of managing capacity risk are coming together to have a profound impact on the social and political structure of the world, as indicated by the Paris Agreement and the recent collapse in energy prices. These trends will deepen into 2016.
2- Cryptocurrencies drive modernization of money (the original virtualization technology)
Money was the original virtualization technology! It decoupled value from goods, simplified commerce, and enabled the service economy. Free from the limitations of physical money, cryptocurrencies can take a fresh approach to simplifying how value (and ultimately trust, in a financial sense) is represented, modified, transferred, and guaranteed in a self-regulated manner. While none of the existing implementations accomplish that, they are getting better understood and the ecosystem built around them will point the way toward a true digital currency.
3- Autonomous tech remains a fantasy, technical complexity is in fleet networks, and all are subordinate to the legal framework
Whether flying, driving, walking, sailing, or swimming, drones and robots of all kinds are increasingly common. Full autonomy will remain a fantasy except for very well defined and constrained use cases. Commercial success favors technologies that aim to augment a human operator. The technology complexity is not in getting one of them to do an acceptable job, but in managing fleets of them as a single network. But everything will be subordinate to an evolving and complex legal framework.
4- Quantum computing moves beyond “is it QC?” to “What can it do?”
A whole new approach to computing (as in, not binary any more), quantum computing is as promising as it is unproven. Quantum computing goes beyond Moore’s law since every quantum bit (qubit) doubles the computational power, similar to the famous wheat and chessboard problem. So the payoff is huge, even though it is, for now, expensive, unproven, and difficult to use. But new players will become more visible, early use cases and gaps will become better defined, new use cases will be identified, and a short stack will emerge to ease programming. This is reminiscent of the early days of computing so a visit to the Computer History Museum would be a good recalibrating experience.
5- The “gig economy” continues to grow as work and labor are better matched
The changing nature of work and traditional jobs received substantial coverage in 2015. The prospect of artificial intelligence that could actually work is causing fears of wholesale elimination of jobs and management layers. On the other hand, employers routinely have difficulty finding talent, and employees routinely have difficulty staying engaged. There is a structural problem here. The “sharing economy” is one approach, albeit legally challenged in the short term. But the freelance and outsourcing approach is alive and well and thriving. In this model, everything is either an activity-sliced project, or a time-sliced process, to be performed by the most suitable internal or external resources. Already, in Silicon Valley, it is common to see people carrying 2-3 business cards as they match their skills and passions to their work and livelihood in a more flexible way than the elusive “permanent” full-time job.
6- Design thinking becomes the new driver of customer-centric business transformation
With the tectonic shifts in technology, demographic, and globalization, companies must transform or else. Design thinking is a good way to bring customer-centricity further into a company and ignite employees’ creativity, going beyond traditional “data driven needs analysis.” What is different this time is the intimate integration of arts and sciences. What remains the same is the sheer difficulty of translating complex user needs to products that are simple but not simplistic, and beautiful yet functional.
7- IoT: if you missed the boat on cloud, you can’t miss it on IoT too
Old guard IT vendors will have the upper hand over new Cloud leaders as they all rush to claim IoT leadership. IoT is where Big Data Analytics, Cognitive Computing, and Machine Learning come together for entirely new ways of managing business processes. In its current (emerging) condition, IoT requires a lot more vertical specialization, professional services, and “solution-selling” than cloud computing did when it was in its relative infancy. This gives traditional big (and even small) IT vendors a chance to drive and define the terms of competition, possibly controlling the choice of cloud/software-stack.
8- Security: Cloud-native, Micro-zones, and brand new strategies
Cybercrime is big business and any organization with digital assets is vulnerable to attack. As Cloud and Mobile weaken IT’s control and IoT adds many more points of vulnerability, new strategies are needed. Cloud-native security technologies will include those that redirect traffic through SaaS-based filters, Micro-Zones to permeate security throughout an app, and brand new approaches to data security.
9- Cloud computing drives further consolidation in hardware
In any value chain, a vendor must decide what value it offers and to whom. With cloud computing, the IT value chain has been disrupted. What used to be a system is now a piece of some cloud somewhere. As the real growth moves to “as-a-service” procurements, there will be fewer but bigger buyers of raw technology who drive hardware design towards scale and commoditization.
10- Composable infrastructure matures, leading to “Data Center as a System”
The computing industry was down the path of hardware partitioning when virtualization took over, and dynamic reconfiguration of hardware resources took a backseat to manipulating software containers. Infrastructure-as-code, composable infrastructure, converged infrastructure, and rack-optimized designs expand that concept. But container reconfiguration is insufficient at scale, and what is needed is hardware reconfiguration across the data center. That is the next frontier and the technologies to enable it are coming.
11- Mobile devices move towards OS-as-a-Service
Mobile devices are now sufficiently capable that new features may or may not be needed by all users and new OS revs often slow down the device. Even with free upgrades and pre-downloaded OS revs, it is hard to make customers upgrade, while power users jailbreak and get the new features on an old OS. Over time, new capabilities will be provided via more modular dynamically loaded OS services, essentially a new class of apps that are deeply integrated into the OS, to be downloaded on demand.
12- Social Media drives the Analytics Frontier
Nowhere are the demands for scale, flexibility and effectiveness for analytics greater than in social media. This is far beyond Web Analytics. The seven largest “populations” in the world are Google, China, India, Facebook, WhatsApp, WeChat and Baidu, in approximately that order, not to mention Amazon, Apple, Samsung, and several others, plus many important commercial and government applications that rely on social media datasets. Searching through such large datasets with very large numbers of images, social commentary, and complex network relationships stresses the analytical algorithms far beyond anything ever seen before. The pace of algorithmic development for data analytics and for machine intelligence will accelerate, increasingly shaped by social media requirements.
13- Technical Debt continues to accumulate, raising the cost of eventual modernization
Legacy modernization will get more attention as micro-services, data-flow, and scale-out elasticity become established. But long-term, software engineering is in dire need of the predictability and maintainability that is associated with other engineering disciplines. That need is not going away and may very well require a wholesale new methodology for programming. In the meantime, technologies that help automate software modernization, or enable modular maintainability, will gain traction.
14- Tools emerge to relieve the DB-DevOps squeeze
The technical and operational burden on developers has been growing. It is not sustainable. NoSQL databases removed the time-delay and complexity of a data schema at the expense of more complex codes, pulling developers closer to data management and persistence issues. DevOps, on the other hand, has pulled developers closer to the actual deployment and operation of apps with the associated networking, resource allocation, and quality-of-service (QoS) issues. This is another “rubber band” that cannot stretch much more. As cloud adoption continues, development, deployment, and operations will become more synchronized enabling more automation.
The idea of a “memory-only architecture” dates back several decades. New super-large memory systems are finally making it possible to hold entire data sets in memory. Combine this with Flash (and other emerging storage-class memory technologies) and you have the recipe for entirely new ways of achieving near-real-time/straight-through processing.
16- Multi-cloud will be used as a single cloud
Small and mid-size public cloud providers will form coalitions around a large market leader to offer enterprise customers the flexibility of cloud without the lock-in and the risk of having a single supplier for a given app. This opens the door for transparently running a single app across multiple public clouds at the same time.
17- Binary compatibility cracks
It’s been years since most app developers needed to know what CPU their app runs on, since they work on the higher levels of a tall software stack. Porting code sill requires time and effort but for elastic/stateless cloud apps, the work is to make sure the software stack is there and works as expected. But the emergence of large cloud providers is changing the dynamics. They have the wherewithal to port any system software to any CPU thus assuring a rich software infrastructure. And they need to differentiate and cut costs. We are already seeing GPUs in cloud offerings and FPGAs embedded in CPUs. We will also see the first examples of special cloud regions based on one or more of ARM, OpenPower, MIPS, and SPARC. Large providers can now offer a usable cloud infrastructure using any hardware that is differentiated and economically viable, free from the requirement of binary compatibility.
OrionX is a team of industry analysts, marketing executives, and demand generation experts. With a stellar reputation in Silicon Valley, OrionX is known for its trusted counsel, command of market forces, technical depth, and original content.
Data is good. Big Data is big, but what does it take to make it good?
All else being equal, it’s better to have data and dismiss it than not have it and miss out. No data, and you’re driving blind. But too much data, and you might as well be driving blind. Sensory overload!
The digitization of human life is just beginning. Information Age is coming and is changing everything. So the best days of Data are ahead of it. We’d better get used to data, lots of it.
Here at OrionX.net, We’ve done several projects to help define business and solution strategies in Big Data, IoT, and Cloud markets. Along the way, some rules and truisms have emerged.
1- Data is cheaper to keep than to delete. Multiple copies, in fact. #NoDelete
In a way, Big Data is enabled by the economics of keeping it around. Nobody dares delete anything because it’s cheap enough to keep, you never know if you’ll need it later, and there may be legal consequences in deleting it.
2- Whatever caused you to collect data will cause you to collect a lot more data. #PlanForScale
Most data collection is focused on ongoing activities so it’s streaming in. Furthermore, as you learn what to do with data, the appetite for even more data usually grows.
If content is king context is kingdom!
3- Big Data systems start small, show promise, go big. #NoMiddle
There are few mid-size Big Data deployments. Once the proof of concept for a project looks promising, they go big and then grow incrementally from there, while spawning new projects.
4- Data must flow to be valuable. Just how valuable is a function of context. #Workflow
Sitting data is an idle asset that is likely depreciating in value. And some contexts are more valuable than others. Think of Big Data as workflow and consider that if content is king, then context is kingdom.
5- Never assume that you know what is cause and what is effect. #ConfirmationBias
In most cases where using Big Data is worth the effort, cause and effect relationships are complex, the data is incomplete, and the users’ biases get in the way.
Reminds me of an epigraph I read years ago: “If there is a will to condemn, the evidence will be found.”
6- The ratio of relevant data to irrelevant data will asymptotically approach zero. #Haystack
One way to say this is: there’s only one needle, and lots of haystack. The more data you collect, the more haystack you’re adding. But the real point here is that for a given context, irrelevant data accumulates faster than relevant data.
7- The ultimate purpose of analysis is synthesis. #Synthetics
When you’re done with analytics, you’re going to want “synthetics”! This is where Machine Learning and Cognitive Computing come in, but also the kind of lateral thinking and connecting-the-dots that only humans seem able to do.
8- Time = Money = Data. There is always a context in which a piece of data is valuable. #ReturnOnData
How valuable is your data and how rapidly does it lose its value? Data is an asset and while it can appreciate in value, it usually depreciates as new data displaces old data and as historical data becomes less likely to be relevant. What is the “interest rate” for your data?
The quality of the insight is a direct function of the quality of data (and the interpretation of that data).
10- Given enough data, you can simultaneously “prove” opposites. #BeautifulMind #Multiverse
The evidence to support any hypothesis will grow with the size of data, asymptotically approaching 100%.
A fully scientific methodology can guard against wrong conclusions, but complexity, (im)proper motivation, malice, or ignorance can lead to invalid conclusions. The more data, the better the odds that one can get confused and make an innocent mistake, cherry-pick to advance a desired belief, or twist the facts to achieve sinister ends. It reminds me of an epigraph I read years ago: “If there is a will to condemn, the evidence will be found.”
11- Most conclusions will be either uninteresting or invalid. Big Data starts with interesting-but-useless and graduates to valid-and-useful. #InsightWins
We live in a world of new media and viral memes where the interesting-but-shallow can trump the insightful-but-boring. Occasionally, something is both interesting and insightful, but long-term, viral witticism will saturate its space and we’ll hopefully get too used to linkbait patterns to be moved. Big Data is about deeper understandings that can improve things beyond one’s immediate mood.
12- Big Data and HPC converge as data volume grows. #Analytics
If you have 200 rows of data, you have a spreadsheet; if you have 2 billion rows, you have HPC! As the size of data grows, you need math and science to make sense of it. Value is increasingly in analytics (and “synthetics” as in item 7 above), which in turn is about math and scientific models. Check out what my colleague Stephen Perrenod wrote in a 2-part series on this topic here and here.
Are these consistent with what you see? Share your insights please.
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
Last week, in Part 1 of this two-part blog, we looked at trends in Big Data and analytics, and started to touch on the relationship with HPC (High Performance Computing). In this week’s blog we take a look at the usage of Big Data in HPC and what commercial and HPC Big Data environments have in common, as well as their differences.
High Performance Computing has been the breeding ground for many important mainstream computing and IT developments, including:
The Web
Cluster computing
Cloud computing
Hi-quality visualization and animation
Parallel computing
and arguably, Big Data itself
Big Data has indeed been a reality in many HPC disciplines for decades, including:
Particle physics
Genomics
Astronomy
Weather and climate modeling
Petroleum seismic processing
Horseshoe Falls (portion of Niagara Falls on Canadian side)
All of these fields and others generate massive amounts of data, which must be cleaned, calibrated, reduced and analyzed in great depth in order to extract knowledge. This might be a new genetic sequence, the identification of a new particle such as the Higgs Boson, the location and characteristics of an oil reservoir, or a more accurate weather forecast. And naturally the data volumes and velocity are growing continually as scientific and engineering instrumentation becomes more advanced.
A recent article, published in the July 2015 issue of the Communications of the ACM, is titled “Exascale computing and Big Data”. Authors Daniel A. Reed and Jack Dongarra note that “scientific discovery and engineering innovation requires unifying traditionally separated high-performance computing and big data analytics”.
(n.b. Exascale is 1000 x Petascale, which in turn is 1000 x Terascale. HPC and Big Data are already well into the Petascale era. Europe, Japan, China and the U.S. have all announced Exascale HPC initiatives spanning the next several years.)
What’s in common between Big Data environments and HPC environments? Both are characterized by racks and racks of commodity x86 systems configured as compute clusters. Both environments have compute system management challenges in terms of power, cooling, reliability and administration, scaling to as many as hundreds of thousands of cores and many Petabytes of data. Both are characterized by large amounts of local node storage, increasing use of flash memory for fast data access and high-bandwidth switches between compute nodes. And both are characterized by use of Linux OS operating systems or flavors of Unix. Open source software is generally favored up through the middleware level.
What’s different? Big Data and analytics uses VMs above the OS, SANs as well as local storage, the Hadoop (parallel) file system, key-value store methods, and a different middleware environment including Map-Reduce, Hive and the like. Higher-level languages (R, Python, Pig Latin) are preferred for development purposes.
HPC uses C, C++, and Fortran traditional compiler development environments, numerical and parallel libraries, batch schedulers and the Lustre parallel file system. And in some cases HPC systems employ accelerator chips such as Nvidia GPUs or Intel Xeon Phi processors, to enhance floating point performance. (Prediction: we’ll start seeing more and more of these used in Big Data analytics as well – http://www.nvidia.com/object/data-science-analytics-database.html).
But in both cases the pipeline is essentially:
Data acquisition -> Data processing -> Model / Simulation -> Analytics -> Results
The analytics must be based on and informed by a model that is attempting to capture the essence of the phenomena being measured and analyzed. There is always a model — it may be simple or complex; it may be implicit or explicit.
Human behavior is highly complex, and every user, every customer, every patient, is unique. As applications become more complex in search of higher accuracy and greater insight, and as compute clusters and data management capabilities become more powerful, the models or assumptions behind the analytics will in turn become more complex and more capable. This will result in more predictive and prescriptive power.
Our general conclusion is that while there are some distinct differences between Big Data and HPC, there are significant commonalities. Big Data is more the province of social sciences and HPC more the province of physical sciences and engineering, but they overlap, and especially so when it comes to the life sciences. Is bioinformatics HPC or Big Data? Yes, both. How about the analysis of clinical trials for new pharmaceuticals? Arguably, both again.
So cross-fertilization and areas of convergence will continue, while each of Big Data and HPC continue to develop new methods appropriate to their specific disciplines. And many of these new methods will crossover to the other area when appropriate.
The National Science Foundation believes in the convergence of Big Data and HPC and is putting $2.4 million of their money into this at the University of Michigan, in support of various applications including climate science, cardiovascular disease and dark matter and dark energy. See:
Stephen Perrenod has lived and worked in Asia, the US, and Europe and possesses business experience across all major geographies in the Asia-Pacific region. He specializes in corporate strategy for market expansion, and cryptocurrency/blockchain on a deep foundation of high performance computing (HPC), cloud computing and big data. He is a prolific blogger and author of a book on cosmology.
Data volumes, velocity, and variety are increasing as consumer devices become more powerful. PCs, smart phones and tablets are the instrumentation, along with the business applications that continually capture user input, usage patterns and transactions. As devices become more powerful each year (each few months!) the generated volumes of data and the speed of data flow both increase concomitantly. And the variety of available applications and usage models for consumer devices is rapidly increasing as well.
Are the Big Data and HPC disciplines converging or diverging?
Holding more and more data in-memory, via in-memory databases and in-memory computing, is becoming increasingly important in Big Data and data management more broadly. HPC has always required very large memories due to both large data volumes and the complexity of the simulation models.
Igauzu Falls: By Mario Roberto Duran Ortiz Mariordo (Own work) CC BY 3.0, via Wikimedia Commons
Volume and Velocity and Variety
As is often pointed out in the Big Data field, it is the analytics that matters. Collecting, classifying and sorting data is a necessary prerequisite. But until a proper analysis is done, one has only expended time, energy and money. Analytics is where the value extraction happens, and that must justify the collection effort.
Applications for Big Data include customer retention, fraud detection, cross-selling, direct marketing, portfolio management, risk management, underwriting, decision support, and algorithmic trading. Industries deploying Big Data applications include telecommunications, retail, finance, insurance, health care, and the pharmaceutical industry.
There are a wide variety of statistical methods and techniques employed in the analytical phase. These can include higher-level AI or machine learning techniques e.g. neural networks, support vector machines, radial basis functions, and nearest neighbor methods. These imply a significant requirement for a large number of floating point operations, which is characteristic of most of HPC.
For one view on this, here is a recent report on InsideHPC.com and video on “Why HPC is so important to AI”
If one has the right back-end applications and systems then it is possible to keep up with the growth in data and perform the deep analytics necessary to extract new insights about customers, their wants and desires, and their behavior and buying patterns. These back-end systems increasingly need to be of the scale of HPC systems in order to stay on top of all of the ever more rapidly incoming data, and to meet the requirement to extract maximum value.
In Part 2 of this blog series, we’ll look at how Big Data and HPC environments differ, and at what they have in common.
Stephen Perrenod has lived and worked in Asia, the US, and Europe and possesses business experience across all major geographies in the Asia-Pacific region. He specializes in corporate strategy for market expansion, and cryptocurrency/blockchain on a deep foundation of high performance computing (HPC), cloud computing and big data. He is a prolific blogger and author of a book on cosmology.
Today, OrionX is celebrating another great milestone as we welcome a new partner to our team! Peter ffoulkes, who was most recently Research Director at 451 Research is joining OrionX.
Thanks to the support of all of you, our clients, our colleagues, our community, OrionX is poised to continue to raise the bar for Strategy, Marketing, and PR.
Here is the press release we issued today which sums it up well.
OrionX Appoints Cloud Computing Industry and Marketing Expert Peter ffoulkes as Partner
Pioneer of a new consulting services model for Strategy, Marketing, and PR, Silicon Valley-based OrionX Bolsters Leadership Team in support of Expanded Services
Menlo Park, CA, September 1, 2015 – Pioneering a new consulting services model for strategy, marketing, and PR, OrionX today announced the appointment of Peter ffoulkes as Partner based in San Francisco. Mr. ffoulkes was most recently Research Director for Cloud Computing and Enterprise Platforms at 451 Research.
“We are thrilled to welcome Peter to OrionX to help serve and grow our expanding client base,” said Shahin Khan, Founding Partner at OrionX. “Peter’s record of excellence, deep understanding of emerging technologies and customer requirements, and his first hand experience leading strategy and marketing organizations perfectly matches the fundamental strategy at OrionX.”
Previously Mr. ffoulkes was Vice President of Marketing at Adaptive Computing, Director of Outbound Marketing at ClearSpeed Technology, Director of Marketing for High-Performance Computing at Sun Microsystems, and Director and Principal Analyst for the Worldwide Workstation Computing program at Gartner Dataquest.
“I am delighted to join OrionX, a group of professionals who are well known across the industry for their depth, breadth, and quality of work,” said Peter ffoulkes. “I am equally excited to work with a customer base that exemplifies the technology leadership and innovation that is transforming the IT industry.”
About OrionX
Mobile, Social, Cloud, and Big Data are redefining B2B technology. OrionX.net is the pioneer of a new model for Strategy, Marketing, and PR. Orion is known for its depth and agility, insightful advice, original content, and its stellar reputation in Silicon Valley. More than 40 technology leaders in virtually every B2B segment have trusted OrionX.net to help set new break-away strategies, ignite brands, and engage customers for increased market share. Visit us at OrionX.net
Shahin is a technology analyst and an active CxO, board member, and advisor. He serves on the board of directors of Wizmo (SaaS) and Massively Parallel Technologies (code modernization) and is an advisor to CollabWorks (future of work). He is co-host of the @HPCpodcast, Mktg_Podcast, and OrionX Download podcast.
If content is king
context is kingdom!