Fast Cache and Storage Pools

EMC probably has internal tools to tell how much of a LUN is in FAST cache, but to the rest of us while it is an interesting question it isn't really useful because you can't influence FAST cache behavior. 

Let's say you were able to find out that none of the LUN in question is in FAST cache.  What can you do about it?  Nothing.  This means that there is something else in the pool which is hotter than the DB LUN you are concerned about.  And in a pool there is only a yes or no to FAST cache for the whole pool. The whole point of FAST cache is that you don't have to guess what portion to allocate to specific LUNs.  The system automatically uses that resource to the best effect on the array.

The only way to fine tune or isolate is to use RAID groups, which have their own management issues.

In your case it is extremely unlikely that putting all the disks in one pool with FAST VP is going to improve the performance of that LUN because it is already on the highest tier of disk available. 

I would first consider what Flare version you are on and upgrade to the latest code revision if possible.  There are some FAST VP and FAST cache improvements in several revisions that may alleviate your problem.  Especially as you probably have DB logs in that pool as well, and small block sequential writes can pollute FAST cache.

If you are on a new enough Flare version (assuming VNX) and are licensed, make sure you have tiering enabled on the pool so that the system will distribute within the tier.  Even though there are not two tiers of disks, the VNX will move chunks around to distribute workload among the private RAID groups of the pool.

Second, you (or someone else) need to do some work to analyze exactly what the problem is.  What is the complaint exactly?  Is it a measured value like latency or it some perceived "thing" that isn't an array problem?

If it is an array problem, is the issue with that LUN, or with workload around it?  Pooled resources are great, but they also have the potential for workloads to be competing with the same physical resources.  A problem can seem like it is with that LUN, but it may be with some other LUN on the system being a bully.

At the end of the day you may need to add resources to the pool.  You can either add more SAS, or you can add an EFD tier.  If you add an EFD tier, you can pin a LUN to it guaranteeing high performance.  However, depending on exactly how large the LUN is this may not be feasible.  And depending on the workload it may be a waste of $$$.

Without knowing specific details it is hard to make a blanket recommendation, but hopefully this information was useful.  I'd suggest engaging an EMC rep or someone similar who can analyze and offer a specific recommendation to improve your performance.