Where traditional virtualization divvies up computer’s resources into many (smaller) virtual machines, clustering seeks to combine them, for failover/reliability, or added performance. (See 3 Ways to Virtualize Applications with Distributed Computing.)
However, traditional clustering approaches can require specific expertise, along with investments in along with storage infrastructure and management. And, points out Steve Conway, Research VP for High-Performance Computing (HPC), IDC, for applications like aerospace, automotive and product design, the relatively low bandwidth of the cluster links results in performance far too slow for today’s fast-paced product design cycles.
Daniel Dern is an independent technology and business writer, who has written one Internet user guide, and thousands of features, reviews and other articles for various technology and business publications, and was the founding Editor-in-Chief at Internet World magazine, and editor of Byte.com. His blog can be found at Trying Technology and he can be reached at [email protected].
Proprietary SMP (Symmetrical MultiProcessing) systems can provide the scale and speed needed, but SMP systems are comparatively more expensive than x86-based clusters.
ScaleMP‘s vSMP Foundation virtualization software lets a company combine – aggregate – commodity x86 boxes into one big system, single virtual symmetric multiprocessing (SMP) system. (Somewhat similar to “channel bonding” used in networking to combine multiple links into one higher-bandwidth connection, an approach colloquially known as “reverse milking machine.”)
Using Intel-based systems, “vSMP Foundation currently aggregates up to 128 x86 systems to create a single system with 4 to over 1,000 processors (that’s over 8,000 cores) and up to 64 TB of shared memory,” according to the company.
This means that a vSMP Foundation environment has only a single OS instance for the entire cluster, doesn’t require clustered storage, or cluster management skills and staff, and can be used as a cost-effective SMP environment for computer- and memory-intensive applications. This makes it an intriguing alternative to conventional clustering and conventional SMP solutions.
Last month ScaleMP announced that their vSMP Foundation “reverse system aggregation” software will now support not just Intel processors, but also AMD processors, specifically, AMD Opteron 6100 Series (“Magny-Cours”) processor as well AMD’s the upcoming 6200 (“Interlagos”) processors.
AMD currently offers Opteron servers with up to four processors. By using vSMP Foundation, an IT shop can aggregate multiple – few or lots! – of AMD-based x86 systems, and create a much larger AMD-based environment. For example, says Shai Fultheim, founder and CEO of ScaleMP, “With 128 four-socket systems, at 16 cores to the socket, you could have an 8,000+ core system.”
(Currently, vSMP Foundation doesn’t let you mix-and-match Intel and AMD machines within a single environment… but that capability may be added in the future.)
Supporting AMD announcement is of interest to compute-density-sensitive data centers, Fultheim notes, because where Intel systems currently need 2U of rack space for a four-socket system, AMD can do it in 1U, and it lets IT do comparison price shopping between the two vendors.
There is a limited release of vSMP Foundation available now, followed by a general availability November 21, 2011.” (This is roughly around when the matching AMD processors will be available.)
