I got confirmed that the test was done in "write through" mode.
So this means that, unless anyone would prove otherwise, indeed its RAID 5 and 6 write performance for SSD RAID:s indeed is 25MB/sec per disk set.
That is complete junk!
I wouldn't even touch that with rubber gloves!
So seems anyone who wants data redundancy needs to implement that in software, as hardware simply not provides that, at least not with any performance whatsoever.
Incredible, this is 2016, I would really have thought that we were past the 25MB/sec mark on top-of-the-line hardware. Anyhow, thanks for the reality check!
Hi everyone - here is an ultra important question:
The RAID 5 and RAID 6 benchmark figures are ultra-bad - the write speed has a 98% overhead (6k IOPS = 25MB/sec), and the read has an 80% overhead (130k IOPS = 500MB/sec), for the set of 8 drives.
Like, really, so ultra catastrophic that I wouldn't even touch that card - and neither its OEM variant "Megaraid SAS3108".
This makes me wonder, however, what settings the benchmark was done with:
The recommended settings are to set "Write cache" to "Write through" mode, to enable "Direct I/O" mode, to enable "Disk's cache" (that is, the physical disks' own cache), and to disable the "Read cache" altogether.
Perhaps absence of running the test with "CacheVault" would be an indirect reason for the crappy numbers - any thoughts on that?
Any completing info on the SSD 4K RW IOPS in RAID 5 and RAID 6 with an actually healthy, proper configuration as described here would be much appreciated!!
Please post your thoughts & actual benchmark results - thanks!!