Power Systems At The Center Of Texas A&M Research
February 3, 2014 Timothy Prickett Morgan
|
Power Systems At The Center Of Texas A&M Research
This is the kind of deal we expect to see more of now that IBM has made it clear to the world that it wants to focus more on its own Power platforms and less on the X86 platforms based on processors from Intel and Advanced Micro Devices. That said, the deal also shows how hard it is to just peddle Power-based machines. Under a broad arrangement with Texas A&M, the ag school that is also the largest research institution in the state, IBM is installing lots of Power-based gear to do a wide variety of research. Michael Dell, being perhaps the most famous alum of the University of Texas, has the inside track on supercomputer deals at that school, which is just down the road in Austin from Dell’s headquarters in Round Rock. This is a pretty big deal for IBM. At the heart of the hybrid system is a BlueGene/Q massively parallel computer system with more than 2,000 nodes that will have 418 teraflops of sustained performance. It will be used to do a mix of commercial and scientific jobs, including analytics and modeling for business-style computing as well as molecular dynamics, protein folding, and organ modeling on the technical computing side. A cluster of 75 PowerLinux 7R2 two-socket Power7+ nodes that are lashed together using 10 Gb/sec Ethernet switches will be set up with IBM’s BigInsights variant of Hadoop as well as its Platform Symphony high-speed Java messaging software, which can be used either by itself or as an adjunct to speed up Hadoop. This cluster will also be equipped with IBM’s Platform LSF job scheduler and the storage on the nodes will run IBM’s General Parallel File System. Another cluster based on IBM’s NextScale hyperscale, vanity-free system designs that debuted last fall, will have 900 nodes and will be used for geosciences and analytics research. Using top-bin twelve-core Xeon E5-2600 v2 parts running at 2.7 GHz, a 900-node cluster would have a peak theoretical performance of around 482 teraflops. If Texas A&M is waiting until later this year to get faster Haswell chips, it could cram even more flops into those nodes. The whole shebang will have 5 petabytes of storage plus 10 terabytes of FlashSystem 820 all-flash arrays, all managed by GPFS. RELATED STORIES What The System x Selloff Means To IBM i Shops What Is IBM Going To Do With Its Systems Business? Will Big Blue Deep Six Its X86 Server Biz? All Your IBM i Base Are Belong To Us Power8 Offers Big Blue And IBM i A Clean Slate Power8 And The Potential Oomph In Midrange And Big Boxes IBM Aims NextScale Hyperscale Boxes At Clouds–And Possibly Power8 Power8 Processor Packs A Twelve-Core Punch–And Then Some IBM Forms OpenPower Consortium, Breathes New Life Into Power Systems And Strategy Execs Switch Roles At Big Blue Server Manufacturing Moved Out Of Rochester, Minnesota IBM Rochester Gets A Piece Of the PureSystems Action A Closer Look At The Flex System Iron IBM Launches Hybrid, Flexible Systems Into The Data Center IBM Reorganization Tucks Systems Under Software IBM’s Plan for an Adjacent, Custom Systems Market international Business (machines)
|
