SGI's NUMAlink provides the fast memory access required to sustain performance of HPC applications. With this technology, the ability to compute complex data sets and other workflows can be performed entirely through memory. This dramatically reduces the time and resources required to manage extremely large data sets.
Low latency, high bandwidth operations are required from today’s HPC applications, and with SGI NUMAlink technology, data can be delivered across cluster nodes at a throughput up to 200 times faster than standard commodity clustering switches. With NUMAlink, data can travel round-trip over a NUMAlink switch in as little as 50 nanoseconds.