[zfs-discuss] [Slightly OT] Anyone using so-called "hybrid" (SSHD) disks?

Ryan How ryan at zbit.net.au
Tue Jan 21 06:15:42 EST 2014


I've only used the original Seagate momentus XT. Never heard of them 
using the flash for write-back caching, even the controllers that pair 
an SSD and a standard HDD don't seem to do that? Anyway, I got 
marginally better read speeds, but writes were the same... It really 
depended on the load, but for my use cache the 4GB flash wasn't enough 
to really do anything much. All in all a bit disappointing...


On 21/01/2014 3:21 AM, Durval Menezes wrote:
> Hello folks,
>
> Some days ago, we rented a real cheap/small server in a data center 
> abroad; its major selling point (besides price) is that it comes 
> equipped with a "hybrid" hard-disk (SSHD, in Seagate's parlance), the 
> Seagate ST500LM000. This disk is supposed to have 8GB of MLC flash 
> internally, and to use it for both read and write-back caching.
>
> So we installed EL6 on it, and started doing some measurements to see 
> whether it's good enough for our case (basically a remote probe). Of 
> course we installed ZoL (0.6.2) and used it to format the file system 
> we will be keeping our data at (even if the machine has only one disk 
> and therefore no redundancy, we at least want to know when our data 
> gets corrupted), and then did a cursory benchmark with it using bonnie++:
>
>     cd /pool2013122601/
>     bonnie++ -u root -r 2048 -f -n 0
>
>         Using uid:0, gid:0.
>
>         Writing intelligently...done
>
>         Rewriting...done
>
>         Reading intelligently...done
>
>         start 'em...done...done...done...done...done...
>
>         Version  1.96       ------Sequential Output------ --Sequential
>         Input- --Random-
>
>         Concurrency   1     -Per Chr- --Block-- -Rewrite- -Per Chr-
>         --Block-- --Seeks--
>
>         Machine  Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec
>         %CP  /sec %CP
>
>         pragmaton  4G           61142  32 31136  23     67249  19
>         100.7   5
>
>         Latency                       89884us     739ms     713ms    
>         323ms
>
>         1.96,1.96,pragmaton,1,1388025063,4G,,,,61142,32,31136,23,,,67249,19,100.7,5,,,,,,,,,,,,,,,,,,,89884us,739ms,,713ms,323ms,,,,,,
>
>
> Just 60-67MBps?!? I sincerely expected better for a system with such a 
> SSD-enhanced disk, as I've been having in excess of 100MB/s from a 
> another ZFS system with a similar but *non* hybrid HD.
>
> Suspecting it could be ZFS (the machine has relatively little CPU 
> power, provided by a single-core Via Nano U2250 running at 1.6GHz, and 
> only 2GB RAM), I reformatted the partition with a more traditional 
> filesystem (ext4) and then retested:
>
>
>     cd /ext4fs
>
> bonnie++ -u root -r 2048 -f -n 0
>               Using uid:0, gid:0.
>               Writing intelligently...done
>               Rewriting...done
>               Reading intelligently...done
>               start 'em...done...done...done...done...done...
>               Version  1.96       ------Sequential Output------ 
> --Sequential Input- --Random-
>               Concurrency   1     -Per Chr- --Block-- -Rewrite- -Per 
> Chr- --Block-- --Seeks--
>               Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec 
> %CP K/sec %CP  /sec %CP
> pragmaton 4G           91993  61  9808   5           11576   3 178.0  11
>               Latency                         429ms 634ms             
>   323ms     320ms
>               
> 1.96,1.96,pragmaton,1,1387837083,4G,,,,91993,61,9808,5,,,11576,3,178.0,11,,,,,,,,,,,,,,,,,,,429ms,634ms,,323ms,320ms,,,,,,
>
> Huh?!? The write speed seems more reasonable (if still mediocre) at 
> ~92MB/s, but less than 12MB/s for *sequential* *reads*?!? WTF?
>
> Reformatting with my a filesystem I'm very familiar with (reiserfs) 
> returned results similar to ext4: mediocre ~90MB/s sequential write 
> speeds, and ridiculously low 10-15MB/s sequential read speeds.
>
> I did some googling for similar issues with this kind of disk, to no 
> avail.
>
> Before I invest some serious time trying to understand what's 
> happening (or, more probably, just ditch this server as unsuitable as 
> I really don't have the leisure right now), I would like to ask: has 
> anyone here ever seen anything similar? Does anyone here have 
> experience with this kind of HD?
>
> Thanks in advance,
> -- 
>    Durval.
>
> To unsubscribe from this group and stop receiving emails from it, send 
> an email to zfs-discuss+unsubscribe at zfsonlinux.org.

To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe at zfsonlinux.org.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.zfsonlinux.org/pipermail/zfs-discuss/attachments/20140121/2e246a77/attachment.html>


More information about the zfs-discuss mailing list