BP’s new supercomputer—now at ‘up to’ 2 petaflops

67,000 CPU system with 536 terabyte RAM shows 267 fold speed-up over 2003 machine.

BP is to up capacity at its high performance computing center in Houston to what it claims is now ‘the largest supercomputing complex for commercial research in the world.’ The new HPC center, scheduled to open mid-2013 at BP’s Westlake Campus will be a hub for processing and managing geologic and seismic data from BP’s worldwide operations. BP’s current HPC center, with a petaflop capacity has maxed out in power and cooling capacity in its current location. A new three-story, 110,000 square foot facility is being built to house the new machine.

We reported on BP’s supercomputer back in our April 2003 issue and it is interesting to see how things have evolved in the past ten years. The 2003 system, built by HP at a cost of $15 million, comprised 259 cluster nodes each with four processors and 32 GB of memory, making for a total of just over 1,000 CPUs, 8 terabytes of RAM and a 7.5 teraflop bandwidth. By 2007, BP’s system had grown to 14,000 cores and 100 teraflops (OITJ May 2007). The 2013 system’s figures are 67,000 CPUs 536 terabytes RAM and a bandwidth of ‘up to’ two petaflops. The result, according to BP is that an imaging job that would have taken 4 years in 2003 will now run in just one day.

In an interview with HPCWire, BP head of supercomputing Keith Gray confirmed that BP still eschews the GPU as a computational accelerator. The new machine is built around a variety of Intel Xeon multi-core CPU nodes from HP and Dell. The network is Arista’s Ethernet and storage system providers include Panasas, IBM, and DataDirect. More from BP.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.