Data Center

July 28, 2017

Data loss is a subject whose time has come! Digital Transformation is upon us and with it the beginnings of a vibrant data economy. Extracting, manipulating, and making sense of the information content from physical assets and business processes is the essence of this new era. With it, come two growing problems: downtime, and data loss.

Downtime

Downtime is interruption, the opposite of business continuity. In a digital enterprise, more than ever, business continuity depends on high availability (HA) and disaster recovery (DR). Both of those depend on fast data replication.

Downtime has been studied for decades. Preventing it is better understood, but it continues to happen from time to time. Sometimes it’s kept quiet and sometimes it’s spectacularly public. But it always causes damage. When a major airline experienced IT failures recently, it not only made the headlines but it cost the company millions of dollars in lost revenue, and perhaps also in damaged brand image. Search online for “IT outage” and you will see sobering stories. Downtime is serious business.

Data Loss

Digital organizations are at the nexus of massive data flowing in and out of them. As they become more common, the value of an organization’s data assets become correspondingly more significant. As a result, one aspect of downtime that is getting more attention is the value of the data that is lost or not captured.

What is the cost to your organization if you lose your payroll, or your customer transactions for a day’s business? These painful types of data loss are harder to quantify and more importantly, ones your company doesn’t want to encounter.

OrionX Research Paper

In our recent research paper, The Cost of Data Loss and How to Avoid It, my colleague Dan Olds examines the drivers and cost categories of data loss. He has also looked at what solutions are available to prevent data loss, and highlights a particularly effective one that is based on Open Source technologies and has been downloaded an incredible 1.7 million times.

Browse through all of our free Constellation papers, or go directly to the full Data Loss paper to learn more.

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 is new in the High Performance Interconnect (HPI) market?

The OrionX Team /

July 14, 2017

Dan and Shahin revisit the high end of the networking market, discuss what’s new, how HPI technologies stack up, and unveil the OrionX rankings of key technologies based on market, customer, and product attributes. Four OrionX research papers on this topic are available here.

The OrionX Team

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.

Does Oracle have the right Cloud Strategy?

The OrionX Team /

June 27, 2017

David Simmons of Oracle and Peter ffoulkes of OrionX are Dan’s guests as they discuss Oracle’s strategy for Cloud in Enterprise IT. Dan Olds and Peter ffoulkes discuss the OrionX blog that instigated this episode.

The OrionX Team

For Enterprise Cloud, Oracle is the Only Choice Grounded in Reality

June 14, 2017

Snapshot of Cloud Computing for Enterprise IT

There is a loudly expressed opinion that ‘the cloud’ is the future of computing and that companies such as Amazon, Google, and Microsoft will dominate the market, replacing established hardware and software vendors such as IBM, Dell, HPE, and Oracle. In some ways, this is true especially for startups, small companies, and perhaps even some midsize companies, but the situation for enterprise level deployments is very different.

For small companies, the use of cloud services minimizes the need for capital expenditure on infrastructure and the recruitment of IT specialists, accelerating their time to production, possibly at the expense of some customization.

By contrast, enterprise level organizations typically have a substantial reliance on established business critical applications. Many of these applications cannot easily be rewritten or re-deployed on public cloud architectures. Sometimes the systems they run on are not based on the x86 architecture, further complicating matters.

Existing enterprise applications, however, can and often need to be integrated with new cloud services to deliver a modern infrastructure as a cloud service while ensuring business continuity. This hybrid approach is an optimum combination of private and public cloud services. It leverages existing data center and infrastructure resources and combines elastic cloud services from external providers when needed. For businesses where regulation and compliance may affect where and how data may be located and accessed, or if large memory scale up data systems may be UNIX-based and are not well-suited or compatible with scale out Intel-based architectures, a hybrid cloud is particularly advantageous.

A Business Solution to the Cloud Problem

Enterprise customers are much more concerned with a business solution to their cloud problem than just a technical solution. Minimizing costly changes is valuable to them as long as the underlying technology has a reasonable and real roadmap. For them, a mixture of private and public cloud can address multiple requirements:

Application availability, without a big cost to port, migrate, and re-deploy
Governance, where ownership, location, and management of resources can be customized for different parts of a distributed cloud application
Modernization, driven by new technologies that enable new capabilities but remain “under the hood”

This approach focuses on business optimization, integrates and consolidates disparate systems, offers a flexible choice of on-premises, managed on-premises, or off-premises cloud.

For enterprise customers, this is a critical transformation. The service provider model is generally accepted as the way forward, but successfully navigating the journey requires the integration of existing business critical systems with new and agile applications. This requires differentiation at a business service level, it is not technology driven but must incorporate both the old and the new while optimizing capital costs and operational margins.

Market Landscape

Cloud computing means fewer but larger buyers of infrastructure technologies. Cloud service providers such as Amazon and Google buy standardized systems in large volumes. Other CPU and infrastructure vendors have doubled down on the capabilities of their architectures to support both legacy and modern systems.

Among established technologies, x86 continues to dominate at the CPU level but is unable to reach a large body of enterprise applications that were written for IBM, HP, and Oracle/Sun. HP has effectively moved to x86 and is not offering its own public cloud service. IBM has formed the OpenPower consortium as it attempts to build an ecosystem and is building its Softlayer cloud service, but relies on application software partners to complete the stack. This leaves Oracle in the unique position of offering chips-to-apps-to-cloud in one place supporting x86 as well as its own SPARC chip (with the SPARC Dedicated Compute Service). As the only UNIX public cloud service available from a major systems or cloud vendor, Oracle has a robust roadmap with everything available on-premises, managed-on-premises, and off-premises.

Comparing ‘Market Hype’ to Reality

For most enterprises, a successful transition to cloud service-based infrastructures is more important than a rapid transition. There are many market research reports that discuss the nature and pace of cloud adoption in ‘the Enterprise’, but one thing they have in common is the limited pace and scope of public cloud adoption. At the end of the day, it is business and workload considerations that drive decisions, not technology.

While the graph (same link as the image) is slightly dated (2015), the trends have not changed and the presented results are still indicative. The common trends demonstrate that dependence upon isolated on-premises systems – even if virtualized – is declining. Private cloud, whether on-premises or hosted is and will be a substantial part of the market, but the dominant and growing sector that seems destined to dominate is hybrid cloud, a combination of on-premises, hosted and public cloud services.

For many enterprises, the challenge is to find a vendor that can offer all of those things with a viable business model. We live in very interesting times when the only global technology vendor that appears to be able to meet all the criteria is Oracle, with a viable alternative to Intel Architecture in the form of SPARC.

Peter is an expert in Cloud Computing, Big Data, and HPC markets, technology trends, and customer requirements which he blends to craft growth strategies and assess opportunities.

New podcast for big ideas in tech

June 14, 2017

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.

Thus was born our first podcast/slidecast series: The OrionX Download™

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.

The rise of AI marks an end to CPU dominated computing

June 6, 2017

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.

________

Note: A variation of this article was originally published in The Register, “The rise of AI marks an end to CPU dominated computing”.

Top-10 Data Center Predictions for 2017

March 20, 2017

A variation of this article was originally published in The Register, “Wow, what an incredible 12 months: 2017’s data center year in review”, Feb 6, 2017.

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.

What do you think?

Deep Learning & Pieces of Eight

July 14, 2016

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?

5 layer network, http://neuralnetworksanddeeplearning.com/chap6.html, CC3.0-BY-NC

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”.

In his blog entry “Why are Eight Bits Enough for Deep Neural Networks”, Pete Warden states that:

“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.

For example nVidia has just announced a system they call the “World’s first Deep Learning Supercomputer in a Box”. nVidia’s GIE inference engine software supports 16 bit floating point representations. Microsoft is pursuing the FPGA route, claiming that FGPA designs provide competitive performance for substantially less power.

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.

Warden, P. 2015 https://petewarden.com/2015/05/23/why-are-eight-bits-enough-for-deep-neural-networks/

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.

The Return of Vector Processing

June 28, 2016

Vector processing was an unexpected topic to emerge from the International Supercomputing Conference (ISC)held last week.

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.

There are other efforts towards building new exascale-class CPUs such, the “Neo processor” that Rex Computing is developing.

Will vector processing re-emerge as a lightweight, high-performing architecture?
What is the likelihood?

Research from OrionX

June 9, 2016

Today, we are announcing the OrionX Constellation™ research framework as part of our strategy services.

Paraphrasing what I wrote in my last blog when we launched 8 packaged solutions:

One way to see OrionX is as a partner that can help you with content, context, and dissemination. OrionX Strategy can create the right content because it conducts deep analysis and provides a solid perspective in growth segments like Cloud, HPC, IoT, Artifical Intelligence, etc. OrionX Marketing can put that content in the right context by aligning it with your business objectives and customer needs. And OrionX PR can build on that with new content and packaging to pursue the right dissemination, whether by doing the PR work itself or by working with your existing PR resources.

Even if you only use one of the services, the results are more effective because strategy is aware of market positioning and communication requirements; marketing benefits from deep industry knowledge and technical depth; and PR can set a reliable trajectory armed with a coherent content strategy.

Constellation™: Research from OrionX

The Constellation research is all about high quality content. That requires the ability to go deep on technologies, market trends, and customer sentiment. Over the past two years, we have built such a capability.

Given the stellar connotations of the name Orion (not to mention our partner Stephen Perrenod’s book and blog about quantum and astro- physics), it is natural that we would see the players in each market segment as stars in a constellation. Constellations interact and evolve, and stars are of varying size and brightness, their respective positions dependent on cosmic (market) forces and your (customers’) point of view. Capturing that reality in a relatively simple framework has been an exciting achievement that we are proud to unveil.

OrionX Constellation Research Model

Types of Data

The OrionX strategy process collects and organizes data in three categories: market, customer, and product. The Constellation process further divides that into two parameters in each category:

Market presence and trends: a vendor’s presence in the segment and its ability to shape or embrace market trends
Customer needs and readiness: a typical customer’s current needs and readiness to adopt a particular product
Product capabilities and roadmap: a product’s existing capabilities and competitive standing as well as its expected enhancements and replacements in the future

This data is then analyzed to identify salient points which are in turn synthesized into the coherent insights that are necessary for a clear understanding of the dynamics at play and effective market conversations and change management.

Types of Report

The resulting content leads to five types of reports that capture the OrionX perspective on a segment. We call these the five Es:

Event: industry milestones and periodic or seasonal topics)
Evolution: historical perspective on a technology or market as a way of defining the segment and understanding the parameters that impact customer sentiment
Environment: the players (stars) in the segment that is in focus
Evaluation: how customer needs have evolved and how they evaluate (or should evaluate) the products in a segment
Excellence: how OrionX analysts see the segment and score each vendor according to six parameters in the three categories of data, two parameters in each category.

OrionX Constellation Decision Tool

Visualizing six distinct parameters, each with its own axis, in three categories cannot be done with the traditional 2×2 or 3×3 diagrams. Indeed, existing models in the industry do quite a good job of providing simple diagrams that rank vendors in a segment.

To simplify this visualization, we have developed the OrionX Constellation Decision Tool. This diagram effectively visualizes the six parameters that capture the core of technology customers’ selection process. (A future blog will drill down into the diagram.)

Sample Reports

While OrionX is not a typical industry analyst firm, we will offer samples of our reports from time to time. We hope they are useful to you and provide a glimpse of our work. They will complement our blog.

As always, thank you for your support and friendship. Please track our progress on social media or contact us for a free consultation or with any comments. We’d love to hear from you.

What is the use of quantum computing without pictures or conversations?

March 7, 2016

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 50^th, 75^th and 85^th 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.

References and attributions:

Paper on What is the Computational Value of Finite Range Tunneling? http://arxiv.org/pdf/1512.02206v4.pdf

Paper on Benchmarking a quantum annealing processor with the time-to-target metric: http://www.dwavesys.com/sites/default/files/ttt_arxiv.pdf

White rabbit from Sir John Tenniel’s illustrations of “Alice’s Adventures in Wonderland” by Lewis Carroll

“White Rabbit” Lyrics copyright Jefferson Airplane/ Grace Slick

Peter ffoulkes – Partner at OrionX.net

Peter is an expert in Cloud Computing, Big Data, and HPC markets, technology trends, and customer requirements which he blends to craft growth strategies and assess opportunities.

Dan Olds Joins OrionX

March 2, 2016

Dan Olds, who founded Gabriel Consulting Group more than a decade ago, has joined OrionX.

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

Silicon Valley-based Strategy, Marketing & PR Consulting Firm Bolsters
Leadership and Expands Services

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

Composition VIII, Wassily Kandinsky, 1923, Shared via #WikiArtApp http://bit.ly/18GCSaj

IoT: The Ultimate Convergence

February 17, 2016

The Internet of Things (IoT) marketplace is expected to run into the trillions of dollars during the next 5 years. ~~Gartner estimates 6.4 billion devices will be part of the Internet of Things this year. And they project $3 trillion of endpoint spending by the year 2020.~~ [updated link] Gartner estimates 5.8 billion IoT endpoints will be in use in 2020 with revenue from endpoint electronics totaling $389 billion globally.

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.

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.

What is your IoT strategy? OrionX can help.

OrionX 2016 Technology Issues and Predictions

The OrionX Team /

January 9, 2016

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.

15- In-memory computing redefines straight-through apps

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.

The OrionX Team

Benchmarks, Schmenchmarks: Big Iron Data and the November 2015 TOP500

November 16, 2015

There has been much discussion in recent years as to the continued relevance of the High Performance Linpack (HPL) benchmark as a valid measure of the performance of the world’s most capable machines, with some supercomputing sites opting out of the TOP500 completely such as National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign that houses the Cray Blue Waters machine. Blue Waters is claimed to be the fastest supercomputer on a university campus, although without an agreed benchmark it is a little hard to verify such a claim easily.

At a superficial level it appears as though any progress in the upper echelons of the TOP500 may have stalled with little significant change in the TOP10 in recent years and just two new machines added to that elite group in the November 2015 list: Trinity at 8.10 Pflop/s and Hazel Hen at 5.64 Pflop/s, both Cray systems increasing the company’s share of the TOP10 machines to 50%. The TOP500 results are available at http://www.top500.org.

A new benchmark – the High Performance Conjugate Gradients (HPCG) – introduced two years ago was designed to be better aligned with modern supercomputing workloads and includes system features such as high performance interconnects, memory systems, and fine grain cooperative threading. The combined pair of results provided by HPCG and HPL can act as bookends on the performance spectrum of any given system. As of July 2015 HPCG had been used to rank about 40 of the top systems on the TOP500 list.

The November TOP500 List: Important Data for Big Iron
Although being complemented by other benchmarks, the HPL-based TOP500 is still the most significant resource for tracking the pace of super computing technology development historically and for the foreseeable future.
The first petascale machine, ‘Roadrunner’, debuted in June 2008, twelve years after the first terascale machine – ASCI Red in 1996. Until just a few years ago 2018 was the target for hitting the exascale capability level. As 2015 comes to its close the first exascale machine seems much more likely to debut in the first half of the next decade and probably later in that window rather than earlier. So where are we with the TOP500 this November, and what can we expect in the next few lists?

Highlights from the November 2015 TOP500 List on performance:
– The total combined performance of all 500 systems had grown to 420 Pflop/s, compared to 361 Pflop/s last July and 309 Pflop/s a year previously, continuing a noticeable slowdown in growth compared to the previous long-term trend.

– Just over 20% of the systems (a total of 104) use accelerator/co-processor technology, up from 90 in July 2015

– The latest list shows 80 petascale systems worldwide making a total of 16% of the TOP500, up from 50 a year earlier, and significantly more than double the number two years earlier. If this trend continues the entire TOP100 systems are likely to be petascale machines by the end of 2016.

So what do we conclude from this? Certainly that the road to exascale is significantly harder than we may have thought, not just from a technology perspective, but even more importantly from a geo-political and commercial perspective. The aggregate performance level of all of the TOP500 machines has just edged slightly over 40% of the HPL metric for an exascale machine.

Most importantly, increasing the number of petascale-capable resources available to scientists, researchers, and other users up to 20% of the entire list will be a significant milestone. From a useful outcome and transformational perspective it is much more important to support advances in science, research and analysis than to ring the bell with the world’s first exascale system on the TOP500 in 2018, 2023 or 2025.

Architecture
As acknowledged by the introduction of HPCG, HPL and the TOP500 performance benchmark are only one part of the HPC performance equation.

The combination of modern multi-core 64 bit CPUs and math accelerators from Nvidia, Intel and others have addressed many of the issues related to computational performance. The focus on bottlenecks has shifted away from computational strength to data-centric and energy issues, which from a performance perspective influence HPL results but are not explicitly measured in the results. However, from an architectural perspective the TOP500 lists still provide insight into the trends in an extremely useful manner.

Observations from the June 2015 TOP500 List on system interconnects:
– After two years of strong growth from 2008 to 2010, InfiniBand-based TOP500 systems plateaued at 40% of the TOP500 while compute performance grew aggressively with the focus on hybrid, accelerated systems.

– The uptick in InfiniBand deployments June 2015 to over 50% of the TOP500 list for the first time does not appear to be the start of a trend, with Gigabit Ethernet systems increasing to over a third (36.4%) of the TOP500, up from 29.4%.

The TOP500: The November List and What’s Coming Soon
Although the TOP10 on the list have shown little change for a few years, especially with the planned upgrade to Tianhe 2 blocked by the US government, there are systems under development in China that are expected to challenge the 100 Pflop/s barrier in the next twelve months. From the USA, Coral initiative is expected to significantly exceed the 100 Pflop/s limit – targeting the 100 to 300 Pflop/s with two machines from IBM and one from Cray, but not before the 2017 time frame. None of these systems are expected to deliver more than one third of exascale capability.

Vendors: A Changing of the Guard
The USA is still the largest consumer of TOP500 systems with 40% of the share by system count, which clearly favors US vendors, but there are definite signs of change on the wind.

Cray is the clear leader in performance with a 24.9 percent share of installed total performance, roughly the same as a year earlier. Each of the other leading US vendors is experiencing declining share with IBM in second place with a 14.9 percent share, down from 26.1 percent last November. Hewlett Packard is third with 12.9 percent, down from 14.5 percent twelve months ago.

These trends are echoed in the system count numbers. HP has the lead in systems and now has 156 (31%), down from 179 systems (36%) twelve months ago. Cray is in second place with 69 systems (13.8%), up from 12.4% last November. IBM now has 45 systems (9%), down from 136 systems (27.2%) a year ago.

On the processor architecture front a total of 445 systems (89%) are now using Intel processors, slightly up from 86.2 percent six months ago, with IBM Power processors now at 26 systems (5.2%), down from 38 systems (7.6%) six month ago.

Geo-Political Considerations
Perhaps the single most interesting insight to come out of the TOP500 has little to do with the benchmark itself, and the geographical distribution data which in itself underscores the importance of the TOP500 lists in tracking trends.

The number of systems in the United States has fallen to the lowest point since the TOP500 list was created in 1993, down to 201 from 231 in July. The European share has fallen to 107 systems compared to 141 on the last list and is now lower than the Asian share, which has risen to 173 systems, up from 107 the previous list and with China nearly tripling the number of systems to 109. In Europe, Germany is the clear leader with 32 systems, followed by France and the UK at 18 systems each. Although the numbers for performance-based distributions vary slightly, the percentage distribution is remarkably similar.

The USA and China divide over 60% of the TOP500 systems between them, with China having half the number and capability of the US-based machines. While there is some discussion about how efficiently the Chinese systems are utilized in comparison to the larger US-based systems, that situation is likely to be temporary and could be addressed by a change in funding models. It is abundantly clear that China is much more serious about increasing funding for HPC systems while US funding levels continue to demonstrate a lack of commitment to maintaining development beyond specific areas such as high energy physics.

TOP500: “I Have Seen The Future and it Isn’t Exascale”
The nature of HPC architectures, workloads and markets are all in flux in the next few years, and an established metric such as the TOP500 is an invaluable tool to monitor the trends. Even though HPL is no longer the only relevant benchmark, the emergence of complementary benchmarks such as HPCG serves to enhance the value of the TOP500, not to diminish it. It will take a while to generate a body of data equivalent to the extensive resource that has been established by the TOP500.

What we can see from the current trends shown in the November 2015 TOP500 list is that although an exascale machine will be built eventually, it almost certainly will not be delivered this decade and possibly not until half way through the next decade. Looking at the current trends it is also very unlikely that the first exascale machine will be built in the USA, and much more likely that it will be built in China.

Over the next few years the TOP500 is likely to remain the best tool we have to monitor geo-political and technology trends in the HPC community. If the adage coined by the Council on Competitiveness, “To out-compute is to out-compete” has any relevance, the TOP500 will likely provide us with a good idea of when the balance of total compute capability will shift from the USA to China.

Peter ffoulkes – Partner at OrionX.net

Peter is an expert in Cloud Computing, Big Data, and HPC markets, technology trends, and customer requirements which he blends to craft growth strategies and assess opportunities.

Big Data and HPC: Convergence? (Part 2 of 2)

http://www.bigdatanews.com/group/bdn-daily-press-releases/forum/topics/3-5m-to-mix-supercomputer-simulations-with-big-data

September 14, 2015

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:

What are your thoughts on the convergence (or not) of Big Data and HPC?

Big Data meets HPC: Convergence? (Part 1 of 2)

September 7, 2015

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.

OpenStack Cloud: “And then there were five!”

September 1, 2015

The IT industry is arguably going through its biggest transformation since the introduction of the IBM PC in the early 1980s, and a cloud service based approach is the likely destination for most enterprise workloads. On August 26th and 27th Mirantis hosted the “second annual” OpenStack Silicon Valley event at the Computer History Museum in Mountain View to a standing room only audience.

Setting the scene for the conference, Alex Freedland, Co-Founder and President of Mirantis, pointed out the market momentum gained by OpenStack in five short years, displacing other potential cloud platforms and propelling it very firmly to being one of just five viable cloud computing platforms for enterprise deployment: Amazon Web Services, Google Cloud Platform, Microsoft Azure, VMware, and the OpenStack Community and Partner Ecosystem.

On the surface this might seem to be a rather bold claim for such a young open source technology offering as OpenStack, but based on my end-user studies at 451 Research prior to joining OrionX, I find the claim entirely credible. It is important to remember that despite all of the hype surrounding cloud computing, most enterprises are still in the early stages of cloud adoption and that it is a complex transition. Although the technology platforms to build a cloud are critical to success, there are many more significant factors that will contribute to success. First and foremost, the shift to a cloud-based approach to delivering IT services is a huge cultural transition and with such transitions also come substantial business risk. This was one of the key findings of my own research, and also the main theme of the talk by James Staten, Chief Strategist of the Cloud and Enterprise division at Microsoft and until recently Vice President & Principal Analyst for cloud computing at Forrester Research.

Another factor in a successful transition is choosing the right partners for technology and business solutions. It is here that OpenStack gains enormous credibility, not so much from the current level of maturity of the OpenStack technology platform but from the massive amount of investment in both money and developers that is being poured into the effort to make OpenStack successful, something that simply cannot be matched by any single company on the planet today. There is still much to be done to turn OpenStack into a true enterprise-grade, production-ready cloud platform but it is also clear that OpenStack is funded for success, which is one very important consideration for any enterprise making a significant bet on the foundation for its future IT services.

There are too many proof points to mention that underpin the level of activity and initiatives supporting the momentum behind OpenStack, but a few recent and noteworthy ones are:

On Monday August 24^th, Mirantis benefited from a $100 million investment led by Intel Capital as part of Intel’s Cloud for All initiative
Also as part of the Cloud for All initiative, on July 23^rd Rackspace and Intel announced the OpenStack Innovation Center, designed to accelerate the testing and deployment of scale-out OpenStack environments.
Even the IRS likes OpenStack! On August 26^th, after a lengthy process, including an initial denial and an appeal, the OpenStack Foundation was granted 501c6 tax exempt status, a clear recognition of the positive impact that the Foundation is having on the cloud computing industry.

So, are we all off to see the wizard, the wonderful wizard of OS [OpenStack]?

Possibly, but not quite so fast! There are still lions and tigers and bears, Oh my! There is still much work to do to bring OpenStack, and indeed all the other credible cloud platforms up to production enterprise deployment standards, and this will take time even if the money and the will is there, which it appears to be for OpenStack. Here are just a few of the initiatives outlined at the OpenStack Silicon Valley summit that point to what is in the pipeline to make OpenStack more enterprise ready:

Core Improvements:

Strategic Investments

The long and winding road

Perhaps the biggest issue of all to be addressed is the current scarcity of OpenStack expertise and developers. They seem to be worth more than their weight in gold or Bitcoins, and they also seem to know it! These are some of the driving forces behind initiatives such as Intel’s Cloud for All, and why a company like Mirantis is gaining recognition for its expertise, and investment that enhances its ability to influence the future of cloud computing.

The journey to “the cloud” is a long and twisty one, and there will be more than one destination, but for now OpenStack seems likely to be one of the five most viable destinations. From here it appears to be more real than it does to be a mirage.

Peter ffoulkes, Partner OrionX

Peter is an expert in Cloud Computing, Big Data, and HPC markets, technology trends, and customer requirements which he blends to craft growth strategies and assess opportunities.