We have an odd problem. Our EMC has a FastCache system that caches on SSD. Unfortunately, it apparently ignores any reads/writes over 128K in size.
So if a Windows server makes R/W calls larger than 128K, it bypasses the FastCache, and slows things down. Note: this is *not* the disk geometry blocksize; I mean the Windows call. We haven't been able to find where you can set that, to not use a blocksize over 96K (as an example). Is that something you can tell Windows? Or maybe a setting for the HBA driver? My boss says he can't find anything definite. (this is a StorPort kind of driver, so it ignores "MaximumSGList" parameter, which is for SCSI reads, I am told) Anyone have any ideas?
