The official press release National Laboratory Lawrence Livermore (Lawrence Livermore National Laboratory) today announced the deployment of a new type of supercomputers. To increase the speed of internal communication in parallel normal operating memory at each node is set block of non-volatile memory NVRAM. Also, instead of hard disks involving solid-state drives. The overall performance of a supercomputer named Catalyst (po-angl. catalyst) is quite modest by today's standards and is at the peak of 150 teraflops. For comparison, the second version of the Chinese supercomputer "Milky Way 2" or Tianhe-2 is theoretically capable of reaching the peak of 54.9 petaflops, which is ten times faster.
Intel recognized that Catalyst - it's more than an experiment, an experiment that can do more where traditional architecture fold, and this sequential calculations for many years. As an example can be DNA or genome sequencing. Officially, Catalyst will also serve to solve biological problems. But in general it comes to handling the so-called big data when you need to analyze unstructured and fragmented . And this is where the new architecture can create a zone of a breakthrough, because the address space to access the RAM for a single DRAM and NVRAM.
Total supercomputer Catalyst contains 324 knots. Each node - two 12-Core Xeon E5-2695v2, which is 7776 cores. Also, each has 128 GB of DRAM. 304 nodes connected in parallel DRAM volatile memory in the form of 800-GB PCI Express SSD Intel 910. Another 12 units are a common pool of data storage in the form of a 3.2 TB based SSD. Related Products :
|