Now that you've invested in the latest SAN technology, I bet you can't wait to calculate your ROI...
Well, one way to increase your return is to start migrating off of direct-attached storage (DAS). Replacing old storage arrays with newer technology requires the execution of a solid migration plan. The successful migration of your existing server install base can consist of hundreds or even thousands of servers running different operating systems.
In many cases, the migration plan calls for a host-based solution. However, if the source and target hardware arrays are identically configured, which means equal LUN sizes on source and target, an array-based solution is possible. Bear in mind, array-based migrations are typically performed by the vendor. So a charge for this service is likely. Host-based migrations can usually be performed by internal resources -- but -- this does not suggest that host-based solutions are easier to perform. In fact, they can become much more complicated since it's possible to re-examine the way the data was originally laid out and change it.
As new storage arrays come to market, the technology in the box may not have been available in the older storage array models. Or perhaps, the technology did not perform up to spec. So therefore, the new features were not seriously exploited. For example, a storage administrator may now consider hardware striping across logical volume sets, instead of software striping across
Part of your migration plan will be to meet with the owner of the data, usually a business unit, and submit to them a detailed plan on what exactly will take place and approximately how long it will take. The plan should include any hardware requirements (I/O boards, HBAs, cables) that will be needed for connectivity. Also, software requirements (host I/O path management, volume manager, OS patches), that will provide path redundancy or are required for support.
Another important point is performance. A common question asked is -– how is sharing a networked storage system, with many other hosts, going to impact my performance? Be prepared to share performance statistics gathered before and after the migration. As part of a service level agreement (SLA) between IT and its customer, performance thresholds would be a point of interest for both parties. Many times the vendor makes certain "promises" on performance and other characteristics that often become part of the agreement.
With these meetings scheduled throughout the enterprise, it's an opportune time to implement a storage management policy that addresses storage usage. In many instances, file system capacities are not nearly at the levels that justify the storage assigned to the host. Reducing file system capacity to a percentage (75% – 80%) of allocated space will inevitably reduce storage cost. One compelling reason to implement a SAN is to fine tune storage utilization. So we are not just migrating data, we're performing intelligent migrations.
Now that we have established a framework to migrate our servers to a SAN, we'll determine what methods are available to us in the next column.
Do you want to see more articles and insights from noted industry observers? Visit the complete Bits & Bytes column library.
For more on DAS to SAN migration, take a look at these links:
Availability during migration
Networked storage to oust DAS, tape to thrive
Storage management software flies high for Air Force lab
About the author
Ernie Gonzalez has served as an independent consultant (GOZCOM INC.) to many of Wall Streets large institutions designing, installing, and managing enterprise UNIX systems. For the past five plus years his focus has been in SAN technology and SAN networking solutions. Contracting with a large storage vendor, he is frequently asked to assist corporate customers leverage their SANs and migrate from past storage solutions. Ernie can be reached for questions or comments at firstname.lastname@example.org.
This was first published in May 2003