An EMC cache is logically divided so that each LUN has at least one "cache allocation" and additional allocations are available to be dynamically assigned as required. The limit of four cache allocations per LUN is intended to prevent saturation of the cache on behalf of any one client of the EMC box. Bear in mind that a single EMC box may hold databases for several different applications running on different servers. This restriction prevents unfair usage of the cache.

The amount of data that can be cached for any one LUN is approximately twice the size of the cache divided by the number of LUNs. Interestingly, no account is taken of the size of the LUN. A small LUN gets just as much space in the cache as a large one. You can use this to your advantage to ensure that log file writes are always fast writes regardless of the intensity of redo generation. You just create a small log file LUN on each mirrored pair and make it slightly smaller than the cache allocation limit. When Oracle generates redo intensively, the EMC box will be able to cache entire log files, and then destage them at its leisure after Oracle has switched into subsequent log file groups. Almost as good as solid state disk!

Regarding your statement about EMC's cache allocations, what exactly does that mean? I assume it means that a limited amount of the data on a disk will be cached, but how does one figure out how much of the disk can be cached?