We recently installed a SAN. This is an XP-512 (HDS 9960) and a software development environment using Rational's ClearCase. All of our tests were looking good until we went into production. Now performance trails off pretty bad under the load. We cannot see any overloading on the servers, the XP, or the network. What would you suggest we look for to isolate the real problem?
Make sure you are balancing your I/O load and your application is tuned properly.
For I/O load, you can provide quality of service for your application by using what HDS calls Priority Access and Cruise Control. These firmware-based utilities will dynamically provision path bandwidth and also check for "hot spots" (frequently accessed partitions) within the storage array and migrate the data to higher performing or less used areas.
This enables QOS for the production applications connected to the SAN. You can also get reports from the subsystem on utilization of cache and back end paths and disks. You can then manually or automatically tune the box based on the analysis.
You can also spread your access across more spindles by using smaller "LDEVS" (Logical formatted units) and creating LUNs from dispersed arrays. This will in effect take the first partition on the first four disks RAIDset and concatenate it with LDEVs from other RAIDsets across more Fibre Channels and array control processors. You can create a LUN that only spans four spindles or a LUN that spans as many as 144 spindles!
If you are unsure of how the LUN is laid out in the back end, check with your assigned HP service professional for more information. If correctly laid out, there should be no reason the storage subsystem should be the cause of performance concerns.
Editor's note: Do you agree with this expert's response? If you have more to share, post it in one of our .bphAaR2qhqA^0@/searchstorage>discussion forums.
This was first published in June 2002