Powering HPC Locally and in the Cloud with 10 GbE
Tom Statchura from Intel attended the IDF 2010 in San Francisco this year where he aided in the demonstration of an ideal scenario for HPC in the cloud—what is most frequently referred to as “bursting” to gain additional capacity. This demo highlighted what is probably the best use case of all for many HPC users—those who have overextended their local cluster capacity and need to be able to burst beyond to secure much-needed resources.
In the demonstration, “10 GbE iWARP was used for the local cluster as the performant low-latency fabric. Mainstream 10 GbE was used in the cloud as it provides the dynamic virtualization and unified networking features required for virtual data centers.
As time marches on and more networking-centered companies are seeing value in the cloud for users of high-performance computing resources, one can only expect that there will be further developments to make the move and use of cloud-based HPC resources less threatening on the security front. Although Intel is not the first company that comes to mind in this arena, the demonstration (not to mention handy imagery provided) shows how there are some changes on the horizon that might be beneficial for those who work between the physical and virtual spaces.
As one might imagine, since it’s Intel’s demo, there is certainly some serious product marketing going on, but it’s nonetheless an important demonstration about the increasing (relative) ease of migrating certain workloads into a public cloud.
Full story at The Server Room