Cheap disks, expensive brains
New Data Center Strategies Newsletter By Andreas M. Antonopoulos, Network World, 08/29/06
A few weeks back I wrote about "the curse of cheap disks" on enterprise organizations: Storage costs have gotten so cheap that the natural response to a request for more storage is to buy more disk. Yet, the cost of managing storage is high and increasing as a percentage of the total cost of ownership. As discussed in Nemertes' just-published "New Data Center" benchmark, storage is growing at a median rate of 22% year-on-year through 2005 and 2006 (predicted to continue through 2007), and many companies top even that growth, reporting growth rates of 100%, 150% and in some cases 300% or more.
How can enterprises keep from increasing their TCOs similarly? A few key strategies stand out as common characteristics among the enterprises we've benchmarked that have been most successful in containing costs:
* First, start a storage SWAT team.
We've said it before, we'll say it again: except for the very smallest enterprises, have at least one IT employee who focuses strictly on storage. Most companies don't do this (and they quite literally pay the price). Companies that distribute responsibility for storage across the mainframe, client-server, desktop and networking folks find they're spending considerably more on storage management than those with dedicated storage groups. That's not surprising, since storage expertise tends to come at a premium (the median unloaded salary for a storage administrator is over $80,000 per year, and can be much higher depending on industry and geography). A simple way to measure the effectiveness of your SWAT team strategy: look at the amount of storage that can be managed by single storage administrator. In our "New Data Center" benchmark, we found almost an order of magnitude variation between the best and worst performers, from less than two terabytes/administrator to more than 20 terabytes/administrator. Companies without SWAT teams invariably do worse than those who have them.
* Second, make sure you effectively deploy storage virtualization software and hardware.
If you're not doing so yet, you aren't alone - more than 70% of benchmarked companies say they aren't yet deploying storage virtualization. But they should, as the most successful firms, do. With storage virtualzation many different storage subsystems are "pooled" into a larger logical (virtual) storage system. The virtual storage pool can then be logically partitioned and allocated to specific applications and servers. Because the logical partitions do not necessarily reside on a specific drive or storage array, they can be reallocated, moved, shrunk or expanded as needed. The benefit? Higher rates of utilization, and much less time required to configure/reconfigure storage subsystems - both of which translate into lower overall costs.
* Finally, make sure you've standardized storage hardware, software, management, policies, and processes.
Some of this final step will result naturally from deploying the previous two steps - but it should also be an explicit goal. In virtually every aspect of IT, standardization is highly correlated with the lowest possible TCO.