Unsolved
This post is more than 5 years old
50 Posts
0
6863
2.5" drive performance specs
Hey Everyone,
EMC documentation lists the 2.5" 10K SAS drives having the same performance specs as the 3.5" 10K drive. I would expect the 2.5" 10K to have similar performance to a 3.5" 15K. I realize EMC may just be giving conservative numbers but we are comparing a 40K with all 2.5" drives to a 40K with 3.5" larger capacity drives so this would be good to know.
If this isn't documented anywhere but EMC engineers have seen performance differences out in the field that would be good to know as well.
seancummins
226 Posts
2
December 17th, 2013 17:00
Garret,
If you compare 10k/2.5" vs 15k/3.5" stats from common drive manufacturer's spec sheets, the OEMs also rate 15k/3.5" drives with better performance specs.
There are two primary metrics to look at:
When you add these two times together, you get the Service Time for the drive -- that is, the average time (in ms) it will take for that drive to complete a single, random small-block IO, with no drive queueing. From the service time (amount of time required for a single IO), we can estimate the number of IO's that the drive can do in a second -- the "rule of thumb" IOPS for that drive.
Consider the 10k/2.5" and 15k/3.5" drives from HGST; you can find their specs at these links:
10k/2.5" -- Ultrastar C10K600 (SAS Hard Drive) | HGST Storage
15k/3.5" -- Ultrastar 15K600 (SAS Hard Drive) | HGST Storage
The 10k/2.5" drive has a latency of 3ms and a seek time of 3.7ms -- for a total service time of 6.7ms.
The 15k/3.5" drive has a latency of 2ms and a seek time of 3.4ms -- for a total service time of 5.4ms.
We can derive the rule of thumb IOPS per drive by dividing 1 second (1000ms) by the service time of the drive.
For the 10k/2.5" drive -- 1000 / 6.7ms = ~150 IOPS
For the 15k/3.5" drive -- 1000 / 5.4ms = ~185 IOPS
Both drives can certainly do more than this, but it will mean that IOs are queuing at the drives, which will result in elongated response times.
You also mentioned that you're comparing a config with 10k/2.5" vs another config with larger 15k/3.5" drives. The fact that the 3.5" drives are larger (capacity-wise) means that, per drive, the IO Density (IOPS per GB) is actually lower on the 15k drives.
If we take those same rule of thumb IOPS above, and calculate the IO density:
For a 300GB/10k/2.5" drive -- 150 / 300 = 0.5 IOPS per GB
For a 600GB/15k/3.5" drive -- 180 / 600 = 0.3 IOPS per GB
So if these two configs you're comparing have the same usable capacity available in the FC/SAS tier, then the 10k tier actually has more performance capability than the 15k tier.
Personally... I'm a fan of 10k/2.5" drives, and I think we'll see their use continue to increase.
Hope that helps...
- Sean
sauravrohilla
859 Posts
0
December 18th, 2013 05:00
Sean, great stuff... thanks
gblack2
50 Posts
0
December 18th, 2013 06:00
Thanks for the response Sean.
In our configs we have all 10K drives so we are comparing a 2.5" 10K to 3.5" 10K for the SAS tier. There would also be the comparison of 2.5" 7200 NL-SAS vs 3.5" 7200 SATA for the SATA tier. Obviously our configurations are quite different to account for the max size of 1TB 2.5" 7200 vs 2TB SATA drives. I definitely agree with you that I like 10K 2.5" drives but I was confused as the specs of a 2.5" 10K and a 3.5" 10K were listed at the same.
Somehow the overall price per TB comes out to about the same. It seems like a no brainer to go with the 2.5" since you can get a larger max configuration of 3200 drives compared to 1920 3.5" drives. Even with the smaller 1TB drives the maximum capacity is larger due to the ammount of additional 2.5" drives that can be used.
seancummins
226 Posts
1
December 18th, 2013 09:00
Garret,
Glad to help. About the 10k/2.5" and 10k/3.5" specs being the same -- not many drive OEMs manufacture 10k/3.5" drives anymore. So when you order a 10k/3.5" drive today, often times what you're getting is a 10k/2.5" drive in a 3.5" carrier... so it may be the same drive, regardless of external form factor.
Thanks,
- Sean
gblack2
50 Posts
0
December 19th, 2013 07:00
Interesting. I have seen that before with Dell but wasn't aware that EMC did that on their VMAX's at all since EMC is pretty strict on what drives they use.