* What storage virtualization should be
As storage and other key data center technologies continue to become more abstract – more virtualized – IT can and should expect to get more out of their systems. IT should expect CPU utilization to increase. IT should expect disk utilization to increase.
Higher resource utilization should either slow the need to add systems or allow some systems to be powered down when their capacity is not needed, decreasing power draws and heat creation for equivalent work or data capacity. Cooling the data center should get easier.
Storage virtualization is poised to proceed to the next level – breaking the still-prevalent direct mapping between storage systems and compute nodes. Most data centers with storage-area networks still run multiple separate SANs. A storage system in that scenario is providing storage only to specific nodes, based on the fact that only they are connected to it. Any layered appliances to do things like encrypt, compress, or de-duplicate data can therefore only serve those nodes.
Creating a single SAN solves some of the problems with this situation, but the underlying issue remains: Storage is treated as a significantly different kind of resource from servers and network appliances.
Ideally, storage nodes would act as service providers on the network, just another component of an enterprise service-oriented architecture. Just as nodes might offer services related to content management (“check item in”) or distribution (“mail item out”) so might storage systems offer services, too: “archive content” or “store highly available” or “store near-line,” for example. In such an environment, the ability of virtualization to conserve resources should increase further.
With storage that is more autonomous on the network, two things should be possible.
One, it should enable even higher utilization – as multiple SANs with sub-optimal utilization become storage services nodes, each striving to maximize its utilization within its defined service parameters (covering things like response times) – and higher utilization should mean lower power consumption.
Two, it should make the use of compression and de-duplication appliances, recreated as service components, easier and more broadly applicable, slowing storage growth requirements and lowering the power consumption growth curve.
There’s a distance to go to make this kind of environment possible, on many fronts. Ideally, for example, it would be built on a data center network serving both data and storage traffic, so that storage service components look like and communicate in the same way as other service components. For NAS – and iSCSI, in some applications – this should already possible. For systems requiring the lossless block delivery of 4G Fibre Channel, though, 10 Gig Ethernet needs to be made more storage-friendly.
There are other reasons to strive towards a service-based data center, and a converged data center network, but the tangible benefits of cooling the place down a bit should not be overlooked.
Editor’s Note: Is your IT organizational structure holding your company back?
We’re assembling a package of stories to be published in Network World looking at the IT organizational structure of the future and we need your help:News Editor Bob Brown
* If your company has recently reorganized in light of new business and technology challenges and opportunities, can you send along “before” and “after” org charts? (Feel free to strip out the names, we’re really just looking for titles. But give us a sense of the size of your IT organization/business.) Can you tell us how the reorganization has worked out?
* If your company is in need of an IT org makeover, what areas are currently most dysfunctional?
* Any advice on what your peers can do to improve their IT organizations right now?
We’re looking to tap into our readership through a technique some call “crowdsourcing.” Knowing we don’t have all the answers, but that our audience probably has a lot of them, we’re hoping to help you help each other.
Please send info to
