I think the picture (from SAN HQ) speaks for itself. Can you anyone tell me why our SSD drives are slower than our 10k drives in average IOPS as well as Read/Write I/O rates? The measurements reflect the last week of activity.
We're on firmware 5.2.1
jakesterpdx
61 Posts
1
May 2nd, 2012 09:00
I'm not sure why you'd say "this looks like an XVS model" when I clearly stated the model was a PS6100XS. The explanation I got from Dell was: "The SAN’s performance is such that the write cache is not being utilized at all because the disks can handle the performance without caching. While you are seeing your SSD drives being outperformed by your 10K, that is only because performance wise and where the capacity lies, the 10K may have more of the data and can handle the access to it." and that makes sense to me.
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
May 1st, 2012 23:00
You'll need to open a case with Dell to gather diags and SANHQ logs to be sure. However, this looks like an XVS model. So there are two RAIDsets in one enclosure. The data is striped across the two RAIDsets. The SAS has larger drivers so more of the data is on the SAS vs. SSD. It's the same thing you see when you have two members in the same pool with different sized drives. The larger member has more of the data so it tends to do more of the overall IO load The XVS is unique in that during a large write burst data will first go to the SSD then get migrated to the SAS drivers. Also, "hot" pages on the SAS will be migrated to the SSD as needed to improve performance. So what you are seeing doesn't surprise me, especially since you aren't at the performance limit of that member.
Also, please upgrade to 5.2.2.
speedstep
9 Legend
9 Legend
•
47K Posts
0
May 2nd, 2012 09:00
This is not a Dell Issue. Pliant Technology 2.5-inch 200GB Lightning LB 200M and
400GB LB 400M Enterprise Flash Drives (EFDs)
Unlike most SSDs on the market, Pliant’s drives are based on SAS (3Gbps) interfaces,
as opposed to the more common Fibre Channel or SATA interfaces.
They claim sustained performance of more than 10,000 I/Os per second (IOPS)
on the LB 400M SSD and 8,000 IOPS on the LB 200M SSD with a block size of 4KB
and a 70/30 read/write mix.
I suspect a wear leveling algorythm is slowing this down because the drives are written
to frequenly and they have run out of good sectors so they are going to die soon.
Many writes is a big problem for ALL SSD's compared to hard drives.
This is also why SSD's slow down over time. Either way This is not Dell's issue.
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
May 2nd, 2012 09:00
That's the MEMBER name not the model number. Only an XVS has the combination of SSD and SAS drives.
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
May 2nd, 2012 10:00
Sorry, you are correct. My bad.
jakesterpdx
61 Posts
0
May 2nd, 2012 10:00
Speed Step, this is a pretty insane answer for our environment. For one thing, we've had the array for about 3 months and they are rated for at least 2 million hours: www.sandisk.com/.../Lightning_3GB_DS.pdf
For another, I'm not complaining that the SSD isn't delivering more than 3Gbps - I'm asking why mine weren't delivering as much as my 10k drives, so the explanation from Dell makes sense: we're not hitting it very hard, and the majority of the data we are hitting happens to be on the 10k drives.
jakesterpdx
61 Posts
0
May 2nd, 2012 10:00
That's simply not true, and a simple Google search would have led you to the PS6100XS:
www.equallogic.com/.../default.aspx
The PS6100XVS has 15k (XV is the roman numeral for 15) drives and SSD, while the PS6100XS has 10k and SSD.