Posted a quick roundup of shots from day two taken from the show floor at ISC’10. The pictures include some close up snaps of gear from IBM’s Blue Waters project, as well as shots from Bull and Supermicro.
Adaptive jazzes Viewpoint for HPC
Last month we did an in-depth story on Adaptive Computing’s new Moab Viewpoint offering, a webby interface for managing large cluster installations. At that time Viewpoint was really aimed at the enterprise and scale-out workload audience, but Adaptive said they were tuning it for an HPC audience.
This week from ISC they’ve announced Moab Viewpoint 1.1 with those changes
In April, Adaptive Computing announced the availability of Moab Viewpoint™ 1.0, a Web portal targeting enterprise customers that require an ideal framework for managing virtual private clouds and the associated server infrastructure. Moab Viewpoint 1.1 will be available in June for HPC users and addresses the specific requirements of HPC workloads.
…New Moab Viewpoint features include:
- New “gadgets” for Moab Viewpoint’s pluggable architecture, such as “My Workload,” provide an intuitive interface for users to manage HPC jobs, services and resources;
- Enhanced administration capabilities allow the identification and management of “troubled workloads” that have been blocked by policy violations, system failures or other issues;
- Cloud management functionality enables users and administrators to implement HPC cloud architectures and services and to submit and manage HPC workloads, including job submission, canceling and re queuing capabilities;
- Policy management functions ensure that cloud service-level agreements are met;
- Flexible billing support allows organizations to customize accounting and cost-allocation systems;
- New reporting, charting and graphing tools provide users and service managers with information needed to optimize use and delivery of services.
The Rich Report: Five Minutes with Cray’s CTO, Steve Scott
Now that ISC has turned 25 and I’m seeing all these familiar faces, I’ve been getting a little nostalgic for the Old Cray Days. And as I look around the show floor at all this capitalism going on, I think about how it all goes back to Seymour Cray, the guy who created the supercomputer industry. So this week it was a great pleasure for me to meet with Cray CTO, Steve Scott.
insideHPC: You know, I started my career at Cray back in 1986, so I just wanted to say how wonderful it is to see the company doing so well and making money again.
Steve Scott
Steve Scott: It is nice. A lot of the credit I think goes to our CEO, Peter Ungaro. He knows the industry inside and out and he put together a management team that got the finances in order. That was a big deal because people would go, “We like what you’re saying, but what about the financial viability of the company?” So we don’t get that any more.
insideHPC: So I’d like to start out by talking about Exascale. I was having a discussion earlier and someone asked which company is going to get there first. It seems to me it’s down to Cray or IBM.
Steve Scott: You certainly can’t count them out. But there’s actually a record to consider. If you go back and look at not the first peak number or linpack result, but the first sustained application Gigaflop, Teraflop, and Petaflop, they were all on Crays. So the first sustained application Gigaflop was on a CRAY Y-MP in 1988. The first Teraflop was in 98 on a CRAY T3E. And the first sustained application Petaflop was in 2008 on Jaguar and the CRAY XT-5. So I’ve gone out on a limb and said publically that the first sustained application Exaflop will be on a Cray in 2018. So that’s our internal target. And this one is going to be harder than the last one. So now that Cascade is kind of in the bag, most of my time is spent thinking about how we get to Exascale.
insideHPC: So what is the superscale user community doing right now to get things going?
Steve Scott: There’s a lot of stuff heating up in the Exaflop race or whatever you want to call it. The DOE Exascale program is getting closer to reality and the DARPA Ubiquitous High Performance Computing program just got off the ground. They’re not specifically targeting Exascale, but UHPC actually grew out of all the study teams that DARPA sponsored: one on hardware, one on software, and one on resilience.
So UHPC is focused on a Petascale in a box, but at power efficiency targets that are good enough for an Exascale. They target 50 sustained Gigaflops per Watt and if you scale that up, that turns into Exascale for 20 Megawatts. So if we can hit the UHPC target, we will be on our way.
Interestingly, the biggest datacenters today are north of 100 Megawatts. That’s the big Internet datacenters, you know Google, Amazon, and Microsoft. So this Exascale system would be 20 Megawatts, which is bigger than any single system today. Jaguar (#1 on TOP500) is the biggest HPC power consumer today at roughly 7 Megawatts, but it’s also the greenest x86 system in the TOP50 in terms of sustained flops per watt.
insideHPC: So if they can hit these targets, the Exascale system would consume roughly three times as much power as Jaguar?
Steve Scott: Yes, and that’s a big factor because you have to consider the budget to run the system and what’s practical from a political perspective. The power budget for a 20 Megawatt system would be roughly $20 Million per year, since power costs about a million dollars per Megawatt year.
insideHPC: It sounds like the next 8 years are going to be very exciting.
Steve Scott: It is fun. And the past five or so years have been pretty interesting as well. We had this huge technology inflection point, you know, with power and delay and transistor counts are going up, a bunch of things came together which said we’ve got to do things completely differently. That’s why we had the whole multicore phenomenon happening.
But when we look out to Exascale, it’s clear that just straight multi-core x86 is not going to get us there. Fundamentally we have to do something different underneath the covers to get the energy efficiency. So that’s why we’re pretty interested these days in looking at forms of accelerated computing. We formed a partnership with NVIDIA and we’re working with them on some future stuff. It’s not going to take over the world overnight, but it’s headed in the right direction.
insideHPC: So what are the most important things you would like potential customers to know about Cray?
Steve Scott: I would say number one is Efficiency. We’re all about power efficiency and system efficiency. Our focus is entirely on sustained computing, sustained results for real scientists. So we work very, very closely with our end users. We have a relationship with them that is not typical. We know their codes and we work on things that don’t make the peak of the machine any better, but things that help them get sustained results. So the fact that the Jaguar system is sitting at number one on the list is nice, but we really don’t care how fast it runs linpack. There are three applications on Jaguar that are achieving over a Petaflop, sustained, and that’s what it’s all about.
I would say the third thing is that Cray now has complete coverage of the HPC space, from the desk-side up to the biggest supercomputers. We’ve got a collaboration with Microsoft where we offer Windows HPC on our low-end, desk-side boxes. And we have now introduced the CX1000, which is a single rack system that uses Intel processors. It has SMP nodes as well as distributed-memory nodes. It also has graphic acceleration, all integrated and ready to run. So HPC is all we do and we can get you started.
LSI storage in France’s new PFLOPS system
Remember that PFLOPS Bull system that the French Atomic Energy Authority powered up late last week (they are billing it as the largest system in Europe, but as I suspected it hadn’t been up long enough to make the June list)? Turns out that it has an LSI storage array attached to it. LSI has been very active this week: they also announced a partnership with Cray that will make LSI the vendor of choice for the new Intel-based CX-1000 midrange super from Cray.
The implementation of LSI storage technology at CEA/DAM is part of a new OEM partnership between LSI and Bull, a leading high performance computing (HPC) solutions provider based in Europe. The CEA, a government-funded technology research organization, initiated a collaborative program with Bull in 2008 designed to extending the data center capability of the Tera 100 supercomputing center. CEA/DAM will utilize the Tera 100 supercomputer for its nuclear weapons simulation program aimed at guaranteeing the reliability of France’s nuclear weapons.
…To meet the needs of CEA’s weapons simulation program, Tera 100 must support the requirements of a broad spectrum of applications by offering a balance between processing power, data throughput and fault tolerance. The Tera 100 storage cluster combines bullx S Series servers and LSI storage technology to deliver 200 GB/s bandwidth to the CEA IT center’s 15 petabyte clustered infrastructure running the LustreTM file system.
CAPS adds Fermi, OpenCL to HMPP workbench
This week the French tools company CAPS-Entreprise announced a major revision to their flagship directive-based compiler technology HMPP (Heterogeneous Multicore Parallel Programming) workbench. The announcement includes news that they are shipping support for Fermi (announced back in November) and have added the ability to generate OpenCL code
CAPS is pleased to announce the availability of an OpenCL code generator within the just released 2.3 version of its HMPP directive-based hybrid compiler. Also, the CUDA back-end generator has been enhanced with Fermi capabilities and this new release brings support for more native compilers with Intel ifort/icc, GNU gcc/gfortran and PGI pgcc/pgfort compilers, enabling developers to freely use their favorite compiler with HMPP 2.3.
Based on GPU programming and tuning directives, HMPP offers an incremental programming model that allows developers with different levels of expertise to fully exploit GPU hardware accelerators in their legacy code.
Sweden adds Cray to national infrastructure, joins list of countries promoting HPC nationally
This week from ISC Cray has announced that Sweden’s Royal Institute of Technology has added a new 93 TFLOPS XT6m (that’s the scaled down version of Cray’s XT6 big boy line of supers) at Sweden’s Royal Institute of Technology (KTH). The purchase is part of a national plan by the Swedish National Infrastructure for Computing (SNIC) broaden access to HPC.
As an aside, this purchase makes Sweden part of a growing trend of countries making significant national commitments to grow their computing infrastructure. That list also includes many of the larger countries in Europe (with the PRACE effort), China (with their new #2 super and a shot at #1 in November), Russia (national HPC program for businesses).
Anyway, from the Cray announement
The new Cray XT6m supercomputer, which uses the new 12-core AMD Opteron(TM) 6100 Series processor, will support scientists in a wide range of disciplines, including computational chemistry, bioinformatics, molecular dynamics, computational fluid dynamics and other areas.
“The Cray XT6m system is an integrated supercomputing solution that will provide Swedish scientists with unprecedented capabilities for increased scientific productivity,” said Dr. Erwin Laure, director of the PDC Center for High Performance Computing. “The new supercomputer will allow users to run larger and more complex problems with higher resolution than was possible with our previous computing resources.”
“We are very excited that PDC/KTH has selected Cray,” said Dr. Ulla Thiel, Cray vice president, Europe. “We are proud to welcome this prestigious institution to the growing community of Cray customers in Europe. The Cray XT6m supercomputer has been a huge success with top European research institutions that need a higher performing and more tightly integrated production system than commodity clusters can offer.”
SGI Announces Altix UV 1000 Customers
Following SGI’s release of Altix UV at SC09 last year in Portland, they’ve now released a few details regarding the first customers of the “large” versions of the machine. Those among the first to benefit from Altix UV include leading scientific research institutions such as Pittsburgh Supercomputing Center (PSC), Oak Ridge National Laboratory, Leibniz Supercomputing Centre (LRZ), University of Cambridge and leading federal integrators that have selected Altix UV for the new architecture.
These shipments of Altix UV are a major milestone for our customers and SGI,” said Mark J. Barrenechea, SGI CEO. “The Altix UV platform is transformative to how customers can deploy the next generation of computing for the world’s most demanding work loads, including traditional HPC, databases, I/O and cyber security.”
According to the release: Altix UV delivers unmatched scalability with up to 2,048 cores with architectural provisioning for up to 262,144 cores, and supports up to 16 terabytes (TB) of global shared memory in a single system image (SSI). Altix UV leverages NUMAlink 5, SGI’s high speed 15GB per second interconnect, and MPI Offload Engine (MOE) acceleration for superior performance.
As one of the first users of Altix UV, we are excited to continue to work with SGI to explore the many new computational capabilities that are now available to the scientific community with its introduction,” said Michael Levine, scientific director at Pittsburgh Supercomputing Center. “The Altix UV product line allows for unmatched scalability and performance for demanding, shared memory applications, and greatly expands the range of HPC computing capabilities.”
For more info, read their full release here.
Voltaire and Platform Computing Join Forces
Voltaire and Platform Computing, today, announced plans to work together on the management of data center resources in virtualized and cloud computing environments. Through the integration of Voltaire Unified Fabric Manager(TM) (UFM) software and Platform LSF, intelligent workload scheduling and resource allocation extends from the application layer all the way down to the network fabric layer.
HPC is a promising early adopter market for cloud computing, with some substantial cloud initiatives under way, mostly on the private cloud side,” said Steve Conway, research vice president in IDC’s High Performance Computing group. “The partnership between Platform and Voltaire aims to provide organizations with the ability to optimize HPC data resources and manage intensive applications across virtualized and cloud computing environments.”
The combination of Platform LSF and Voltaire UFM software over InfiniBand enables our mutual customers to experience new levels of performance and efficiency for managing HPC data center virtualization and cloud computing technologies,” said Tripp Purvis, vice president, business development, Platform Computing. “The integration of Platform LSF with Voltaire UFM software extends Platform LSF’s ability to optimize the allocation of resources in the data center across the entire network for more efficient IT operations.”
For more info, read their full release here.
Infiniband Goes Wild on the Top500
According to a release by Mellanox today at ISC, Infiniband products have experienced record growth on Top500 list as of late. Frankly, I believe them. From their release: As reported in the 35th edition of the TOP500 list, the growth of Mellanox InfiniBand-connected systems increased to more than 41 percent of the TOP500. Mellanox InfiniBand enables the highest system efficiency and utilization, up to 96 percent.
Mellanox’s end-to-end 40Gb/s connectivity provides the most scalable, reliable, and highest performance solution for demanding compute applications around the world delivering not only the highest return on investment, but also the lowest power consumption,” said Eyal Waldman, president, chairman and CEO of Mellanox Technologies. “We are experiencing the growth of InfiniBand in the high-performance computing markets and are now seeing its adoption in additional markets such as financial, database, virtualized data centers, Web 2.0 and Cloud Computing. InfiniBand has passed the early stage of adoption as a server and storage interconnect and is now entering into a growing number of mainstream vertical markets.”
Advancements in multicore CPU architectures and the rise in GPU computing are placing greater demands on server and storage interconnects for higher bandwidth and lower application latencies,” said Jie Wu, Research Director, Technical Computing at IDC. “The results on the TOP500 list are a clear indication that Mellanox InfiniBand solutions have been gaining traction by many supercomputer buyers. New technology solutions developed by Mellanox, such as application offloads and GPU cluster efficiency enhancements, will continue to address the market demand for higher performance and improved efficiency.”
For more info, read their full release here.
Microsoft and Novell Team up for HPC
….don’t worry, I was shaking my head when I read this too. However, Microsoft and Novell kicked off another partnership this week during ISC. This time its true to form in the HPC market. The partnership was designed to bring down the overall TCO of HPC ownership and improve immediate system management features.
Companies around the world are realizing the benefits of our joint interoperable cross-platform technical solutions,” said Ted MacLean, general manager for Strategic Partnerships and Licensing at Microsoft. “The fact that we’re able to address a real need in the HPC market is evidenced by the number of licenses we’ve issued. These solutions, coupled with Novell’s proven technical support programs, make it easier than ever for our customers to have confidence.”
According to Microsoft, this HPC solution represents another joint technical initiative that delivers value to customers and partners. Improving datacenter optimization for customers all over the globe, this partnership continues to broaden its joint technical scope.
For more info, read their full release here.
SGI Announces Hybrid Petaflop Architecture
SGI announced rough details today about a new hybrid computing platform that will enable 1PF per cabinet performance. Wow. This is some serious performance.
According to the release, the design leverages silicon from possibly NVIDIA, ATI and Tilera. It doesn’t go into specifics about which ones and/or what the specifics might look like inside of an SGI system. However, the technologies might be seen inside an SGI by the end of 2010.
SGI shows its R&D strength again,” said Steve Conway, IDC research vice president for high performance computing. “Following closely on the heels of the Altix UV series, SGI has introduced a highly dense, scalable technology designed for strong sustained performance on very demanding HPC applications.”
The press release is very sparse for good technical meat.
For more info, read the full release here.
Overheard at ISC, Day 2
Our Overheard at ISC coverage continues with these little jewels submitted by our readers:
Guy #1: So above the booth it says: Innovate, Insight, Solve. Is insight a verb?
Guy #2: I think they meant to say Incest.
***************
Gal: I hear that Chauncey Gardener is working for BP Global PR.
**************
Guy digging himself deeper (to woman he can’t remember): Please don’t feel bad. I don’t even know who my own nieces are and they make fun of me all the time.
**************
Guy: Last thing I remember saying was, “This better not get back to my girlfriend in Portland.”
Intel part of new exascale research lab in Germany
At ISC on Monday Intel, Forschungszentrum Jülich, and ParTec announced that they’ve signed a multi-year agreement to create a new ExaCluster Laboratory located on the campus of the Research Center in Jülich, Germany
(left to right) Achim Bachem (Jülich), Kirk Skaugen (Intel), Hugo R. Falter(ParTec) at the signing ceremony.
The new lab will be a private/public collaboration and will explore the key challenges of building computing systems with a thousand times the performance of today’s fastest supercomputers.
The ExaCluster Laboratory will initially employ about a dozen researchers and is expected to approximately triple its employment over time.
“The Forschungszentrum Jülich has taken a leading role in driving high-performance computing research in Europe,” said Kirk Skaugen, Vice President and General Manager of Intel’s Data Centre Group. “We have chosen to work with Forschungszentrum Jülich and the ParTec Cluster Competence GmbH, because of their strong history of innovation in the area of high-performance computing.”
The ExaCluster Laboratory will conduct research into current challenges in systems management software for large heterogeneous super computer systems, with a view to scaling this software to reach exaFLOPS performance. This will include research on open exascale runtime system software, software tools and simulation software.
ISC10 Day One Debrief
Prior to getting terribly deep into the second day’s worth of news and scuttlebutt, I wanted to take a quick moment and put together a list of ‘honorable mentions’ for day one. As with the Supercomputing show, day one at ISC is full of interesting prognosticators, predictions, tutorials and the opening gala. I was privy to a few interesting moments and conversations throughout yesterday’s organized chaos and gala opening. Here are a few juicy tidbits.
- Steve Wallach and Tom Sterling discussing technology. These two people have collectively contributed more than their fair share of charisma and insight into the history of HPC. Quite a privilege to sit and hold court with these characters. Don’t worry, its a New York thing.
- China is fast moving up the Top500 list. LOTS of scuttlebutt on the exhibition floor that they are seriously aiming at the top spot. I’m interested to see the response from the various competitors. There isn’t much time between now and November.
- Less is more. ISC has a lot less glitz and glam. The setups are relatively simple with less ‘in-your-face’ ad campaigns. Its actually quite nice.
- Next-IO: Had a long conversation with my neighbors from Austin, TX. This might be something to keep an eye on. Virtualized I/O is not a new concept, but I believe their switched PCIe ASIC might be the best implementation I’ve seen. Very cool stuff.
- IBM is here with a big board. No. A HUGE board. What would I do if I owned my own fab? Well, build a motherboard the size of a Volkswagon. Why not, right? I’ll try to get some shots with relative scale and post later this week.
So far, ISC has been a real joy. Hamburg is a lovely city [and the beer is incredible]. Check in early and often with insideHPC for all the goings-on for this year’s ISC show.
University of Frankfurt Goes with Bright Cluster Manager
Bright Computing announced this morning that the University of Frankfurt has selected their Bright Cluster Manager management stack to support the new machine at the Goethe University’s Hessian high-performance computer organization (HHLR-GU). The machine, called “LOEWE-CSC”, has 20,784 processor cores plus 772 GPGPU hardware accelerators. Overall theoretical peak stands around 599 TF [double precision].
This supercomputer represents a new era for high-performance computing at the University of Frankfurt. The LOEWE-CSC supercomputer will allow us to run a large range of scientific applications at an unprecedented scale, but also at a ground-breaking performance-to-power-consumption ratio,” said professor Volker Lindenstruth, Chair of HPC Architecture at the Goethe University. “We are pleased to run Bright Cluster Manager on the LOEWE-CSC system because it has already proven itself on another TOP500-class supercomputer at our university.”
The machine will be built by ClusterVision using the Bright Cluster Manager stack. At the heart of the cluster will be 386 units of 2U Supermicro GPU-optimised Twin servers with two hot-plug nodes each, 1544 AMD Opteron 12-core “Magny-Cours” CPUs, 772 ATI Radeon GPGPUs, and QDR Mellanox InfiniBand. Check out their full release here.
