By: Dzezik

Dzezik — Tue, 22 Mar 2011 23:05:20 +0000

in similar test of LSI MegaRAID 9260-8i (https://www.thessdreview.com/our-reviews/lsi-megaraid-9260-8i-review-storage-manager-software-fastpath-and-cachecade/) You have used LSI MegaRAID FastPath software which is no elese than disabling cache for SSD both for writing and reading. Here You are trying to show performance without cache but it shows that LSISAS2108 cache is faster than 8 SSD. is or is not? it depends. new unused SSD can handle 30k IOPS read write but used one only about 3000 IOPS in write (this is a rule almost for all SSD that real no TRIM performance drops 10 times after few hours of heavy load and overprivisioning an GC helps only a little). LSISAS2108 can handle 80k IOPS in normal mode (cache read write) and 240k IOPs in FastPAth mode (no cache). It shows that disabling cache make sense for reading (8 x 30k =240k > 80k) but it is not a good idea to disable write caceh for SSD: 8 x 3000 = 24k < 80k !!!.
LSI takes $160 for this software, but You can do better manual optimization for SSD without is just disabling cache for reading on controller used for SSD.
this software make sense when You attach SSD and HDD to the same controller, then CacheCade is right choice for You.
but all this stuff (RAID controllers with cache) means nothing when You use ordinary HBA and setup ZFS with some SSDs as cache and log devices.

Comments on: Areca 1880ix-12 PCIe 6G SAS RAID Controller Review SSD IN RAID 0 – PCMark Vantage and Conclusions

By: Dzezik