On 01/27/2015 11:20 AM, Yasha Karant wrote:
> There are hybrid disk drives that have a "small" SSD flash memory along with the regular spinning media, typically to serve as a cache for lower latency.  Does anyone have long term experience with these units?  Most flash memory is not designed for the long term repeated read/write/erase cycles of a primary disk drive -- how are these holding up?  Is the flash configuration totally transparent to the Linux file systems and formatting operations?
>
> Yasha Karant

Disclaimer: Since I work for the US Fed Gov, I am not endorsing or disapproving any products, just sharing my experience.

My experience with the Seagate 2TB's ST2000DX001 is they're getting SMART warnings of pending/offline sectors in a typical desktop setting after just a few months of 24/7 use, where other drives have lived for years.  My last failure locked up the Dell BIOS and the system wouldn't boot unless you hit F12 to boot the other drive.  I think I've only deployed about five drives and at least three are in the 'dead' pile already.  Reviews weren't bad when I bought, but it's pretty ugly now with 25% out of 200 being 1-egg ratings: http://www.newegg.com/Product/Product.aspx?Item=N82E16822178380&SortField=0&SummaryType=0&PageSize=10&SelectedRating=-1&VideoOnlyMark=False&IsFeedbackTab=true#scrollFullInfo

I don't think the SSD is to blame, it's the mechanicals.

Again, I use them as replacements a RAID1 configuration, and sadly, haven't really done much in the way of benchmarking them, but I suspect if you reboot the system a few times, the boot times will come down, but agreed, as mentioned, 8GB of SSD (mostly read) cache doesn't help much unless your read/write patterns are fairly restricted.

What I have found that *does* work well is RAID1 mirroring with SSD and a traditional "spinning rust" hard drive and using mdadm to mark the partitions on the HD as "--write-mostly" (fail the devices, remove them, then re-add them with --write-mostly).  You get speed, redundancy, and low cost.

As I manage roughly 170 systems, most desktops which have space for two drives, I cannot emphasize how nice it is to not have a system go down due to a common drive failure.  I get to replace them on a schedule that works best for me and the user, not when they're screaming "I'm dead in the water!"