Z4 with Microsemi 2100-4i4e - connection for 4-drive RAID

[wny2016]

We just received an off-lease Z4 with a Microsemi 2100-4i4e RAID controller. Looks like the refurbisher removed all but one drive and reformatted, Windows 10, etc. That controller is 12Gbps.

My initial inclination was to install 4 x Samsung EVO 870 SSDs, RAID1 as that is what we use in many of our machines.

All of the "popularly priced" SSDs are only 6Gbps, AFAIK. But the controller is SAS compatible, and 12Gbps enterprise-level SAS drives are plentiful and pretty inexpensive.

Which way to go? I realize that these days, spinning-disk technology is considered to be old hat...

TIA for your suggestions!

[DGroves]

raiding ssd's on this controller for a speed increase (raid 0) is pointless, to do so for data redundancy (raid 1) is worthwhile

unless you require large amounts of storage, a mech SAS drive is not preferred nowadays also hardware raid levels 5/6 are not recommended for large capacity arrays due to rebuild time and possible errors during a restore

note that raid 1 is mirroring, which only uses 2 drives, raid 10 is raid 0 + raid 1 combined for speed and data redundancy and might be worthwhile on arrays smaller than 6TB

so if you do go the mech 12 gbps SAS route i  also suggest trying raid 10

if the array is greater than 6TB then i don't recommend a dedicated raid controller, use a software only solution like "ZFS"

[wny2016]

Thank you for your detailed explanation. An array of NOS 10K/12mbps SAS mechanical drives, 900MB apiece, would be extremely inexpensive to set up, as those new drives (with a mfr date of maybe 2015 or so) cost less than $40 apiece = $160. With four of those, the array would have 1.8Tb available for use. My main concern is really not speed, but reliability. 

An array of four 1Tb SSD drives, 6Mbps, costs about $70 apiece for the drives = $280. So not much cost difference between the two choices.

I worry though, about SSD reliability, as we had another HP Z-series workstation where we installed a pair of brand new Samsung 860 1Tb drives as RAID1 using the Intel RST controller. One drive failed with about 5 hours of light use, had to RMA late last year. There are other reports in the tech sector of current Samsung SSDs failing at higher than expected rates, as well.  Enterprise-level SAS 10K mechanical drives, which I have used often in the past, hardly ever failed.

[DGroves · Accepted solution]

hardware failures are a fact of life, currently the failure rate between mec drives and ssd's are about equal

i will also point out that the most common failure point on a ssd is the controller chip not the nand chips

 raid is not a substitution for a data backup,... a proper backup plan for a small amount of data is usually something that runs at night and backs up all files to something like a external drive however this has a flaw in that if a fire or something else happens to the site where the computer is.....................well you get the idea right?

which is why a remote off-site backup may be necessary

last, any backup needs to be tested and verified working!! i can't tell you how many calls we get from companies that have a drive failure or corrupted file server and then discover their backups  were not functional...........

[wny2016]

Yes, I get the backup concepts. We use a large NAS for daily incremental timed backups, and output the important data to an external SATA drive that gets taken offsite at the end of the day. Can't have too many redundant copies. Also thinking of adding a cloud storage option, as Synology has some pretty cheap and easily configured plans.