I'm quite partial to traditional 2.5" SATA SSD's; I have about 30
servers with SAS/SATA slots running either Kingston or Sandisk 3Gb/s
SSD's in 480Gb capacities. My home SAN runs 20 x 1TB SSD's (also
Kingston) and if I rummage through my gig bag, I'll probably find half a
dozen 1tb M2 SSD's in sata/usb3 cases. After about 6+ years in 24-7 run
capacity (yes, servers do get rebooted and wiped/rebuilt of course) in
effectively webserver / asset server /db server setups - meaning lots of
writes, I have yet to have a 2.5" 3GB/s SATA ssd fail.
Conversely, those "ultra-awesome" Crucial Micron M2 SSD modules I have
had fail on 4 separate occasions - all of them within "warranty," and
Crucial was not able/willing to RMA any of them - completely lousy
customer service, which tempted me to just "buy and replace" through
amazon (no I didn't, morally incorrect, but tempting). I also have some
of the hybrids (both early Hitachi, whatever Apple was using in the
early mac pro tubes) - many of those have failed, so I avoid hybrids
like the plague, even if that new Fire series from Seagate is touted as
the next best thing....for full transparency, I do have another SAN
shelf with 24 1TB 2.5" traditional spindles (because it's an SAS-only
shelf without interposers) that has been a solid performer for a long
time - probably up 5+ years now, the only time off was moving the server
racks and power failures. It's a Netapp shelf, so somewhat surprising
that it has held up so well (nothing to do with the drives however).
Which just goes to show that mileage may vary wildly. I could have a
dozen drives go out within 5 minutes of hitting send, or not. But for
power savings and speed*, and not having to worry about what happens if
a server is mounted directly on top of the UPS stack, or how the drives
get transported, SSD media is a benefit in my book.
(* - my server installs have shown to run faster against the default SAS
72GB slow drives that my servers come with - some folks have shown that
SSDs can be slower than fast HD's with specific testing, and that stable
platters consume less continuous power than idle SSDs during initial
writing. My power bill tells a different story.)
Cheers,
Ted.
_______________________________________________
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users
