Consolidation and Scalability in the Enterprise

Virtualization has long promised what were once considered two opposing goals: increased scalability and reduced hardware. But now that we have established that you can, in fact, do more with less, the race is on to see how far both envelopes can be pushed.

By Arthur Cole | Posted May 26, 2010
Print ArticleEmail Article
  • Share on Facebook
  • Share on Twitter
  • Share on LinkedIn
Virtualization has long promised what were once considered two opposing goals: increased scalability and reduced hardware.

But now that we have established that you can, in fact, do more with less, the race is on to see how far both envelopes can be pushed.

So it's no surprise that the recent development of a 128-node server aggregate platform from ScaleMP has drawn a substantial amount of interest. First off, the new vSMP Foundation package offers a dramatic increase in scalability over the previous 16-node version, simultaneously boosting the address space from 4 TB to 16 TB. The system also accommodates up to 128 threads per node, which means you could conceivably have a single SMP box working 16,384 threads at one time. How's that for scalability?

But wait, there's more. The system also has the ability to stack virtual machines on top of other virtual machines. The way the company describes it, you can place a KVM or Xen VM together under the single-system management view on the vSMP Foundation aggregator. And since it allows all physical servers in the cluster to be managed as one unit, you can more closely align your management stack with the way the cluster appears to available software and applications.

Any way you shake it, though, consolidation is a tricky endeavor, producing an increasingly complex environment for application and data functionality. So it's probably a good idea to have a firm grasp on the impact any consolidation effort will have on existing pathways. Groups like CiRBA are adding new analytical tools to data intelligence platforms with an eye on the fact that multiple components of single applications are often spread across a range of servers. CiRBA 3.5 maps out things like TCP/IP connections and business service membership parameters so that current dependencies are not interrupted during the consolidation phase.

If done right, consolidation will prove its worth on the bottom line. The U.S. House of Representatives recently wrapped up a two-year consolidation effort that has allowed the organization to shave its energy bill by about $2,000 per day. With a pre-consolidation CPU utilization rate of about 5 percent, the House IT staff was able to trim its 85 Windows servers down to eight and cut an additional 35 UNIX machines to 12. The next step is to hopefully cut down on the number of individual servers maintained by each congressional office.

Fortunately, most enterprises have the luxury of all their users being on the same team, so consolidation is not hampered by concerns of sharing physical resources with the enemy. But it still remains a laborious process, albeit one that can produce even greater rewards with increased levels of scalability.

Comment and Contribute
(Maximum characters: 1200). You have
characters left.
Get the Latest Scoop with Enterprise Networking Planet Newsletter