November 18, 2008
This week Sun is announcing new Constellation wins at Sandia and Aachen University (200 TFLOPS by 2010), which come on the heels of large announcements at Forschungszentrum Jülich (the 207 TFLOPS Sun/Bull joint venture) slated for installation in 2009, the Korean Institute of Science and Technology Information system that will ultimately grow to 300 TFLOPS, and a system in Canada at the Ontario Cancer Biomarker Network. These systems represent a mix of Suns own processors and, in the case of the Sandia, Jülich, and Aachen systems, Sun's next generation Intel blade.
Bjorn Andersson, director of HPC marketing for Sun, spent some time with me before the conference talking about the booth and where the company is headed next in HPC. With recent wins, and the successful Ranger deployment at TACC featuring heavily in the company's conference activities, Sun feels it has a great deal of momentum moving into 2009 for its HPC business. A major push this week is to build on that momentum, and to start to grow awareness of storage as a major part of Sun's HPC strategy. "More than 40 percent of systems in the Top100 use Lustre," says Andersson of the parallel file storage technology acquired last year by the company, "and that includes 7 of the top 10 systems on the list. Beyond that, 54 percent of the 50 most powerful systems on the Top500 list use Sun's StorageTek tape storage silos."
Sun is also previewing the blades that will drive some of those major new Constellation wins on the show floor. Andersson says Sun will have working Nehalem-based blades as well as blades based on AMD's new Opteron (Shanghai) processor in the booth. Conference goers will be also get a peek at the next generation of Constellation -- with double the cores and double the compute nodes of the original Constellation, the new Magnum InfiniBand switch, the "Glacier" cooling door, and storage flash arrays.
Sun will also be talking about a new direction for its HPC business as it grabs for a piece of the much-sought-after mid- to low-end HPC market. Sun is packaging its storage and cluster offerings into pre-packaged offerings to make it easier for customers to get started in HPC. The Sun Storage Cluster is a Lustre-based modular storage system and, with a reference to the Sun Storage 7000 family announced last week, Andersson says that "the Sun Storage Cluster is almost like a Lustre-based storage appliance." The system combines Sun Fire servers and data servers and can be configured from 48 terabytes to multiple petabytes, and to scale performance from 1 GB per second to more than 100 GB per second.
The Sun Compute Cluster is a pre-configured HPC cluster built from Sun Fire or Sun Blade servers. Sun says that the Compute Cluster brings Constellation's technology down to the departmental level. The systems come tested and pre-loaded with Sun's HPC stack and either an InfiniBand or GigE network with either 1U rackmount or Blade 6000 servers loaded in anywhere from 1 to 8 racks (for between 30 and 256 nodes), ready for customers to install their application software.
The system will be offered in vanilla and specialized configurations aimed at particular applications areas, initially computer aided engineering and financial services. When asked about the specific differences between the CAE and financial services offerings, Andersson pointed to differences in the interconnect (10GbE for CAE, 1 GbE for financial services) and the operating systems. While the CAE cluster personality uses the Sun HPC Linux stack, in financial services the choice is Solaris for its container technology.
While customers will be able to get started with the Compute Cluster on the Web, Sun anticipates needing to connect directly or through a partner with customers before the made the final purchase. Of course, the pre-configured compute cluster market already has plenty of entrants, with similar offerings from HP, Rackable, Appro, SGI, and others. The most significant value that any entrant can bring to this market right now is a message that connects mid-range mom-and-pop engineering analysis firms and medium-sized businesses with the technology that HPC has to offer them.
Rounding out the big strategic thrusts -- storage, compute, and ease of entry for customers -- is a raft of product announcements. Lustre 1.8, HPC ClusterTools 8.1, HPC Software Linux Edition 1.1, HPC Software Solaris Developer Edition Beta 1, Sun Studio Express 11/08, and Shared Visualization Software 1.1.1 are all getting air time in the booth this week.
If all of that isn't enough to get you into Sun's booth this week Sun has also trucked in the Java Chopper, a Sun-themed custom motorcycle. The bike is accompanied by a photographer providing souvenir photos so that you can prove once and for all that, no, supercomputing is not geeky, thank you very much, at your kid's next school career day. On second thought, maybe you should just let it hang on your cubical wall in silent tribute to HPC bad boys everywhere.
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?