In my last post I talked about the situations where it is not advisable to use EMC FAST VP and FAST Cache. In this post I will talk about an example where you should not employ both of them.
Microsoft Exchange 2010 was designed to use large, slow drives, and minimizes access to the physical disks. As a result, FAST Cache is only useful if very high levels of performance are required. Jetstress, used in testing during Exchange deployment, has poor data locality, so FAST Cache is not likely to provide any deterministic performance improvement with Exchange 2010.
Background Database Maintenance, which is a regular part of Exchange implementations, causes pollution of FAST VP statistics collection and ranking. Homogeneous Pools, or Traditional LUNs, will not exhibit this effect, and are recommended. The use of “Highest Tier Available” data placement for Exchange data may reduce the effect of Background Database Maintenance on FAST VP LUNs.
The use of Thin Pool LUNs should be avoided with Exchange 2010. If Thick Pool LUNs are used, users should be aware that Jetstress causes data to be allocated to LUNs unevenly, causing initial LUN performance to be poorer than that for other LUNs. This could cause Jetstress to report a failure.
To work around this, EMC engineering has developed a utility called “SOAPTool” which forces even distribution of the data. If using Thick Pool LUNs, use the SOAPTool utility to ensure optimal performance. Alternatively, Traditional (RAID Group) LUNs may be used instead.