Book Review – Contemporary High Performance Computing

Yesterday I received my copy of Jeff Vetter’s new book, Contemporary High Performance Computing and while I have not read all 702 pages from cover to cover yet, this book is a great way to quickly come up to speed on the current landscape of HPC if you don’t have time to fly around the world and visit the 21 HPC centers highlighted in the book.

Saying this is Jeff’s book is a bit misleading, as he was only the editor (and of course the one with the idea). Most of the book is composed of overviews of 18 top HPC sites and 3 HPC/Cloud initiatives. Jeff does give an introduction to trends in HPC, followed by a chapter on the HPC Challenge benchmark suite by Jack Dongarra and Piotr Luszczek and a chapter on the Green 500 list.

With the DoD’s TI-13 RFP due this month, vendors bidding on this year’s proposal might want to spring for 1-day delivery on Amazon and read Chapter 6, Supercomputing in the DoD High Performance Computing Modernization Program by John West.

Of course, I particularly enjoyed reading Chapter 7 on the Keeneland Heterogeneous GPU system and Chapter 20 on TSUBAME2.0, both HP supercomputers. Chapter 11, Supercomputing at Moscow State University was a treasure trove of Russian history starting with the Strela supercomputer installed in 1956.

Other chapters cover such notable HPC systems as China’s Tianhe-1A system, of course Blue Waters and Titan get their own chapters, as do systems at NASA, SDSC, PSC, and the French Tera 100 system. Now a question I have to ask Jeff the next time I talk to him is why Tera 100 was selected for the first systems chapter? Could it be because Paris has the best croissants?

If you are at all interested in modern day supercomputing, I highly recommend this book. Given how fast the technology in HPC evolves, it should make equally interesting history to pick up and look at 10 years from now!

Posted in Uncategorized | Leave a comment

HP Insight CMU Adds Support for Moonshot Servers

After yesterday’s big coming out party for HP’s new Moonshot servers in New York City, lets take a look at how HP Insight Cluster Management Utility (CMU) can be used to manage the Moonshot 1500 chassis and Moonshot servers.

Originally developed for the provisioning, monitoring, and management needs of HPC clusters, and used by many systems on the Top500 list, CMU has rapidly expanded its capabilities in recent years and is now the tool of choice for many HP hyperscale customers, not only for HPC, but also for other scale-out workloads like Hadoop. Look closely at the screen shot below and you can see how CMU’s 3-D Timeview display correlates Hadoop Map and Reduce progress with CPU, disk, and other system metrics.

As shown in the top figure, the HP Moonshot 1500 chassis management module provides standard IPMI and serial console capabilities for each Moonshot server. But with up to 180 servers in a chassis (future capability) and 1800 servers in a rack, a tool like CMU quickly becomes indispensable. Each Moonshot server cartridge, along with the Moonshot chassis cooling and power sub-systems and the chassis fabric, communicate with a shared chassis management module via satellite controllers (SC). The chassis management module works in conjunction with the satellite controllers to implement HP’s widely used iLO (integrated lights-out) management functionality in a scalable manner.

In addition to full support for the HP Moonshot 1500 chassis, CMU v7.1 introduces several important new features including:

  • efficient and scalable storage and visualization of all numerical monitored metrics for archived user groups
  • numerous GUI improvements including a significantly more efficient 3D Timeview display
  • a new automated installation tool
  • API improvements
  • Back-end improvements

    Finally, CMU v7.1 also introduces new support for monitoring Intel Xeon Phi co-processors.

    With all these improvements, CMU still manages to stay true to its core, and remains a simple, easy to use, sys-admin focused tool for provisioning, monitoring, and managing large clusters, which these days are not only limited to HPC clusters but clusters running Hadoop and now a new breed of Moonshot software defined servers.

  • Posted in Uncategorized

    GPU Technology Conference Wrap-Up

    It has been a busy week at Nvidia’s GPU Technology Conference, mostly filled with customer meetings for me. In this wrapup blog I’ll focus on some personal observations and try to extract a few trends rather than try to document the week’s product announcements that have been well covered elsewhere.

    I’ve been coming to this conference for many years, and each year there is notably greater attendance by a broader audience. An Nvidia employee shared pre-registration numbers with me, showing increases over last year, and the show floor seemed considerably more crowded than even the pre-registration numbers hinted at. Despite growth in attendance, the conference continues to stay true to it’s technical roots, with many excellent technical sessions. Walking through the show floor to peruse my competitors’ wares, booth conversations seemed dominated by technical discussions, vs lookie loos. There was also a marked increase in displays and presentations outside of the core workstation and server markets that domindated the first and second phases of GPU adoption. Specifically in automotive technology, Audi, BMW, Tesla, and several exotic manufacturers had their vehicles with Nvidia-powered infotainment systems spread around the conference center with several offering test drives to conference attendees (sorry, the McLaren and Lamborghini were not available for test drives).

    As expected, Nvidia shared a few high level details on future high-end Tesla GPUs used in servers. Stacked memory in the next gen + 1 Volta processor seemed to draw the most press coverage and interest, although stacked memory in reality is an industry trend that many are working on so while interesting to see Nvidia throw their weight behind, this isn’t in itself a novel idea. Personally I was much more excited about some of the announcements Nvidia made on the mobile front with their Tegra line, specifically full CUDA and OpenGL support in the Logan or Tegra-5 chip. Promising a Logan chip “the size of dime”, Nvidia is one of several ARM vendors coupling interesting computational add-ons to their ARM cores. While not aimed at CUDA workloads, TI’s KeyStone II ARM + DSP chip takes a similar approach. While traditional server-focused processors like Xeon or Tesla are not going away, it is hard to argue against the sheer numbers presented by the mobile market, which will require billions of processors a year vs millions a year for the server space. With the right packaging, some of these mobile chips are likely to find their way into server platforms for specific workloads.

    The market share growth of any hardware platform is really a three step cycle, both in the consumer/mobile market and the enterprise/server market. Step 1: end users tend to pick their hardware platform based more on the applications available on the platform than because of specific hardware features. Step 2: software developers, in turn, pick the hardware platforms they are going to write to and support based on the platform volume which represents their potential market. Step 3: the more apps are developed for a hardware platform, the more users the platform attracts. Now go to step 1 and repeat.

    Nvidia’s vision of a CUDA platform stretching from dime sized Logan chips to super high end Volta stacked memory powered supercomputers maximizes the potential market size for CUDA application developers. And while no server vendor has yet to announce a Logan based server, as more and more CUDA applications are developed, there are likely to be interesting server use cases for Logan and other mobile processors. Already, HP with Project Moonshot has discussed servers using ARM and Atom mobile processors. As the three step process I outlined above continues to play out for CUDA, I expect future GPU Technology Conferences will have many vendors displaying not only Tesla based servers but interesting Tegra based servers as well.

    As much as I enjoy thinking through the business model impacts of technology announcements, perhaps the most rewarding and refreshing parts of attending a show like the GPU Technology Conference is talking to the end users and developers using the technology. Stanford University’s use of GPU technology to advance Alzheimer’s Research, which was highlighted at the show, is one great example. Yesterday evening I spent at least an hour, I lost track of the time, with an Nvidia researcher and one of the lead developers of Gromacs discussing how the Gromacs team has steadily increased performance by enabling key parts of the code, first on single GPUs and most recently for multi-GPU performance. The developer, based in Sweden, is a scientist by training and today his work is completely supported by government and private grants. “I participated in a few start-ups” he told me, “but I decided I much more enjoyed being a scientist and working on Gromacs than starting businesses, not that there is anything wrong with the latter”. Adding, “living in Sweden, we have a social system that lets me do my job and worry about being a scientist rather than worrying about how to pay for healthcare or sending my kids to college”. Maybe some non-technology lessons there for other countries.

    So once again a great week and a great conference. While some observers of the information technology market may clamor that there isn’t a lot of innovation going on these days in how enterprises process their payroll and accounts receivables, and one report after another expounds on the death of the traditional PC, Nvidia seems to be pointing their research and development straight at the heart of the mobile and high performance computing markets which continue to be a hotbed of innovation and growth.

    Posted in Uncategorized

    HP ElitePad Review

    I’m always excited when a new gadget arrives to test and the HP ElitePad was no exception. Designed with the enterprise user and corporate IT department in mind, the ElitePad also promises to deliver the style and user experience that consumers demand. Using a company provided laptop or tablet shouldn’t be a cause of embarrassment, be it going through airport security or around the family dinner table. So lets see how the ElitePad lives up to the promise and how it compares to the HP Envy X2 I reviewed earlier.

    First the hardware specs. The ElitePad has a 10.1″ diagonal display with a Corning Gorilla Glass 2 screen and 1280 x 800 wide angle display. It comes with two built-in cameras, a rear facing 8.0 MP ideal for high resolution stills and a front-facing 1080p HD camera ideal for Skype or other video conferencing tools. The processor is a dual core 1.80 GHz Intel Atom processor and storage is provided by a 64G SSD drive. More on accessories and connectivity later.

    While the shipping box was light enough, the size seemed rather big for a 10″ tablet.

    Opening up the box doesn’t yield a lot of clues to what is included in addition to the tablet itself.

    But alas, a bit of unwrapping yields the answer.

    In addition to a lightweight AC adapter and the ubiquitous video dongle, the box contains a sturdy dock with a multitude of connections. In fact, after just a few hours use, it seems like the functionality of the dock is going to be one of the key selling points of the ElitePad for enterprise users.

    The back of the dock has a standard VGA connector, an HDMI connector, power, and ethernet.

    The side of the dock has two USB connections and a headphone jack. I plugged my HP wireless keyboard adapter into the USB and the ElitePad immediately recognized it without having to load any additional drivers. I don’t care how good the on-screen keyboard is, power users will want a real keyboard for extended use.

    The ElitePad comes with a full version of Windows 8, not the slimmed down Windows RT designed for tablets, and that should come in handy for running the myriad of corporate applications expecting a full-blown Windows OS that an enterprise user needs. The setup process did however identify a few places where Windows 8 was not 100% tablet ready. Plugging the ElitePad into the dock sits the tablet horizontally. While I had no problem turning my head to read every last word of the license screen:

    Filling out the registration screen using on the on-screen keyboard didn’t work too well with the screen horizontal and I had to remove the tablet from the dock to view it vertically.

    Luckily a few screens farther into the setup process, Windows 8 finally recognized it was on a tablet and the screen correctly displayed horizontally in the dock.

    Running standard Windows 8 no doubt helped HP’s own corporate IT department create the ElitePad configuration screen, which walked me through setting up secure access to HP’s corporate web site, loading a full version of Office Professional, and other internal HP setup.

    One of the other goodies included in with the ElitePad is 50GB of free storage on Box.net. I wonder if Box is using <a href="http://hpcloud.comHP Cloud for their back end storage?

    So far so good. Together with the HP EliteBook 9470m laptop I started using a few weeks ago, the two weigh less than the EliteBook 8440 I recently replaced so traveling should be no problem. Even more exciting, my laptop bag even has room for one of the upcoming HP Moonshot servers we will be launching soon. So check back here for more stories on all three.

    Posted in Uncategorized | 3 Comments

    Nvida Kepler GPU Momentum

    Ever since the HP ProLiant SL powered TSUBAME2.0 system debuted at #4 on the November 2010 Top500 list, the ProLiant SL platform has been one of the world’s leading purpose-built servers for GPU computing. Since Nvdia’s latest Kepler line of GPUs has been shipping for several months now, I thought it would be good to give an update on how HP customers are adopting Kepler.

    HP currently offers three different Kepler GPUs, the K10, K20, and K20X. These GPUs are available in HP’s ProLiant SL250s and ProLiant SL270s servers. First, it is important to understand Nvidia’s naming convention. The K10 is not an entry level into the Kepler line but essentially a special purpose GPU optimized for certain types of applications requiring primarily single precision (vs double precision) performance. One of the largest adopters of the K10 has been the oil & gas industry, where one of the primary workloads, Reverse Time Migration, or RTM, is dominated by single precision math. So it should come as no surprise that the K10 has been a great seller with our worldwide oil and gas customers.

    The K20, on the other hand, is Nvidia’s mainstream GPU and if you have used earlier Nvidia GPUs like the M2070 or M2090, you will love the K20. HP has shipped several mid-size (by server count) GPU clusters including a 264 node SL250s system to USC with 528 K20 GPUs and a cluster of over 100 SL250s servers each also with two K20 GPUs to Clemson. The K20X GPU is similar to the K20, only with more memory and more GPU cores and we currently have multiple customers testing K20X systems.

    Before we shipped USC their 264 node cluster, HP’s High Performance Computing benchmarking team had a few days to use the system in our factory to get a couple of Linpack runs completed. While I won’t share the exact numbers, the system would easily rank in the Top50 on the current Top500 list. That is a pretty amazing testament to the power of Nvidia’s Kepler technology when with only 264 servers you can build one of the 50 fastest supercomputers in the world (given the semi-annual performance improvements in the Top500 list, I won’t make any projections as to where the USC system will land in the next Top500 list to be published this coming June). Compared to earlier generations of GPUs, we have seen significant efficiency improvements with Kepler (measured performance / peak theoretical performance) not only on Linpack but also on a wide range of real world applications.

    Expect many more updates on Kepler at Nvidia’s upcoming GPU Technology Conference, March 18-21 in San Jose, California.

    Posted in Uncategorized | 2 Comments

    History Lessons for Cloud Computing

    2013 will no doubt see continued adoption of Cloud Computing for both consumer and enterprise use. No matter if you are a cloud builder or a cloud user, it is worth looking back over the last 30 years or so of computing and networking to see what historical lessons we can learn. Clouds, by their nature, bring together many of the different technologies that have evolved over the last 30 years and thus many of the same lessons learned in other information technology sectors might have relevance to the future of cloud computing.

    The original NetWare product, introduced in 1983, was one of the first technologies to bring networking to PCs. Back then, it was unheard of for a PC to come standard with a network card much less any networking software. The IBM PC, introduced two years earlier in 1981, was still in it’s infancy but any business that had two or more PCs was a candidate for networking.

    Networked minicomputers had been around for a bit longer, including the DECnet protocol introduced in 1975. However, it wasn’t until 1982 that DECnet evolved into a seven-layer OSI-compliant stack. The early 1980′s was definitely a heyday for the fast evolving world of computer networking.

    Proprietary networking standards like NetWare or DECnet provided generally better performance and greater functionality than Ethernet in the early 1980s. While Ethernet predated even DECnet, having been developed between 1973 and 1974 at Xerox PARC, the first commercial Ethernet implementations didn’t become available until the early 1980′s and were not standardized until about 1985. Of course today, NetWare, DECnet, and virtually every other proprietary networking technology has ceased general use and virtually every PC, laptop, tablet, and smartphone uses Ethernet and related standards to connect to the Internet.

    The lesson for Cloud Computing from the network world is that standards, especially open standards, matter. Of course at a basic level, every cloud today uses some standards, for instance Ethernet, to connect to networks. But just like the ’80s and ’90s saw competing operating systems in both the consumer and commercial space, Cloud Computing implementations today run the gamut from proprietary to open. Lets take a look at where along this spectrum some of the major cloud players are headed.

    One of the oldest cloud computing players, Amazon Web Services or AWS for short, has used its own proprietary software platform since launching in 2002. No one else was thinking about cloud computing in 2002 and there simply didn’t exist software or open standards for cloud computing so Amazon built their own. If you want to build your own cloud, you can’t buy the AWS software from Amazon, however, you can run an AWS-compatible cloud using software from Eucalyptus a startup out of UC Santa Barbara. Eucalyptus, who’s CEO Martin Mickos is no stranger to open source, having run open source database company MySQL and then selling it to Sun Microsystems for $1B. Following the successful MySQL model, Eucalyptus makes a base version of their software available for free and open source, and makes money by selling and supporting commercial versions. Leave it to Martin and team to figure out how to make money using an open source business model making software compatible with one of the largest proprietary clouds.

    AWS originally focused on enterprise offerings, services initially used by startups like photo sharing site SmugMug. Today, countless large commercial companies run at least bits and pieces of their enterprise on AWS, if not their core business. While AWS is generally reliable, it’s widespread use leads to major publicity when there are outages, like when Netflix users lost service this past Christmas eve. Today, AWS also offers consumer services like the Amazon Cloud Player that lets you store and play your music.

    Speaking of music, Apple, the company that reshaped the music industry with iTunes and iPODs, is a major player in the consumer cloud business, where it follows its traditional proprietary software model. It is perhaps only because they have a limited presence with enterprise services that Apple isn’t more often thought of as a top cloud company, but iCloud quickly became one of the largest clouds as Apple enabled users to move their music, video, and other content onto the cloud instead of being trapped inside your music player. But even Apple can’t make all things it touches golden. Apple Maps, heavily dependent on services from the Apple cloud, was recently launched with less than favorable comparisons to Google Maps. In fact, AllThingsD technology writer Liz Gannes recently wrote that in 2012 she basically stopped using Apple apps on her phone. She still has an iPhone, she just uses apps from Google and others.

    Perhaps one of the most used cloud services, across both enterprise and consumer clouds, is still email. Gmail, launched by Google in 2004, Yahoo! Mail launched in 1997, and Hotmail, launched in 1996 and later acquired by Microsoft are three of the top cloud based email services. No doubt what continues to make email so popular is the ubiquitous standardization. Email can be seamlessly exchanged not only across the three cloud email systems above, but across 1000′s of different consumer and enterprise email systems. Think if it was that easy to move you music from Apple iTunes to Amazon Cloud Player or Google Music Player.

    While Microsoft had a slow start in cloud computing except for some consumer services like Hotmail, it’s recent focus leaves no doubt as to Microsoft’s serious plans for the cloud. Microsoft’s cloud platform, Windows Azure wasn’t even launched as a developer platform until October 2008. But what they lacked in head start they have certainly attempted to make up for ever since. Windows Azure is an enterprise cloud platform and no doubt shares technologies with Microsoft consumer cloud services. A great example of Microsoft linking their cloud platform with their consumer products is the new Windows 8 network login. Buy an HP Envy x2 or dozens of other Windows 8 laptops, and one of the first screens you get when you startup the laptop prompts you to create a Windows network login if you don’t already have one. Once you create your Windows network login, all key Windows configurations are stored in the Windows cloud and you get instant access to your desktop configuration whenever you use your network login to sign into any Windows 8 PC. While some devices like Google’s Chromebook have similar features, the Windows 8 network login is one of the most complete I have seen.

    On the enterprise side, Windows Azure has made up for a late start as well. As one would expect, Windows Azure is primarily focused on cloud services for Windows, but Microsoft has made a number of moves to attract non-Windows users to the platform. After first going down the road of developing their own “big data” applications, Microsoft relented to the dominance of Hadoop and now supports a 100% Hadoop compatible service on Azure. I find it interesting that Eucalyptus takes the proprietary AWS and make a compatible open source implementation, while Microsoft takes open source Hadoop and implements a compatible proprietary implementation on Azure. Someone will no doubt right an article 10 years from now comparing the success of Azure vs AWS.

    To attract another big crowd of open source fans, HPC users, Microsoft recently announced a number of improvements to their Azure HPC services including new compute instances supporting up to 120 GB of memory and 16 cores per node, along with a high speed Infiniband interconnect. Kudos to Microsoft for their Hadoop and HPC services, I can’t wait to see what Azure rolls out in 2013.

    Another late comer to the public cloud space is HP. Sometimes not being first has its advantages, and HP’s timing allowed the HP Cloud to take advantage of the OpenStack standard upon which the HP Cloud is built. While OpenStack is by no means the only Cloud standard, it does seem to be gaining momentum and taking the lead in the standards based cloud space. Over time, HP’s cloud business is one of the few that seems well position to play a major role in both consumer and enterprise environments, building on HP’s consumer and enterprise businesses as those customers move to the cloud.

    Unless you live in a mythical land like Mordor, there is not likely to be one cloud to rule them all. Public, private, or hybrid, consumer or enterprise focused, the next few years are likely to see many different cloud platforms come and go. But just like the proliferation of networking technologies 30 years ago, today’s cloud landscape is filled with offerings that no doubt will not make it through the decade and it would be well heeded for cloud builders and cloud users to study history. By the end of the decade there isn’t likely to be a lot of remaining clouds that are not based on or at least compatible with the open source cloud standards that are taking shape today.

    Update #1
    Congratulations to James Cuff on his move to CycleComputing as their new CTO. CycleComputing specializes in software that lets you run large HPC and other types of jobs on clouds like AWS. Speak of a company that could benefit from standards in the cloud space. James will be a great addition to the CycleComputing team and they definitely just moved up a notch in my cloud computing companies to watch.

    Posted in Cloud Computing | 1 Comment

    2012 in Review – Courtesy WordPress.com

    An interesting year-end summary of my blog prepared by WordPress. A good example of the insights you can get from “big data” analytics.

    The WordPress.com stats helper monkeys prepared a 2012 annual report for this blog.

    Here’s an excerpt:

    4,329 films were submitted to the 2012 Cannes Film Festival. This blog had 21,000 views in 2012. If each view were a film, this blog would power 5 Film Festivals

    Click here to see the complete report.

    Posted in Uncategorized