VMWare ESXi Datastore latency - acceptible levels

I know this is a grey area as there is usually more involved to it.

We have a HP BL460 G1 with 4x 1TB 7,200 SAS Drives in RAID 1+0 (512MB BBWC at 50/50% read/write sharing) at present but may need to push it up to RAID 5 for the extra storage.

Currently the datastore latency is as follows:
Write Latency: Maximum 8ms, average 1.1ms
Read latency: Maximum 40ms, average 6.5ms

The I/O's on the server aren't that high, but what could I expect if I changed from RAID 1+0 to RAID 5? I'm assuming +25% increase in maximum latency - would that sound close enough?
LVL 6
kiwistagAsked:
Who is Participating?
 
asavenerConnect With a Mentor Commented:
Well, you encounter a penalty for the RAID level, but you get a benefit depending on how many spindles are in your RAID array.  If you throw enough spindles at the problem, your IOPS go up, and your latency goes down.
0
 
asavenerCommented:
7200 rpm drives are typically SATA, not SAS.  Can you verify the hardware?

Then tell me what you're proposing as your end state, and we can get an pretty good estimate of the IOPS you'll see.
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
Change the BBWC configuration to 75 % Write and 25 % Read.

Performance in RAID 10, is high for some operationns in Read and Write, changing to RAID 5 would decrease the IOPS.

IOPS calc here will give you some idea of figures, but inserting some numbers for RAID 10 and RAID 5.

http://www.wmarow.com/strcalc/
0
Upgrade your Question Security!

Your question, your audience. Choose who sees your identity—and your question—with question security.

 
kiwistagAuthor Commented:
asavener: They are Seagate ST91000640SS (CONSTELLATION.2 1TB SAS 64MB) disks. Not your standard HP disks..
Although we do have a HP Gen8 DL360 with 4x 1TB 'Midline SAS' (7,200RPM) drives also here.

Andrew: If the latency on Reads is higher than writes, would it not be better to have the cache optimised for reads (75% Read)? I could be viewing it from the wrong angle mind you.
I do understand that RAID 1+0 offers the best performance (with redundancy) and anticipated the IOPS loss.

Thanks, that calculator has helped too.
This is with the basis of 65% read.
RAID 1+0:
average random IOPS: 226.83
bandwidth (MiB/s): 1.77

RAID 5:
average random IOPS: 149.38
bandwidth (MiB/s): 1.17
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)Connect With a Mentor VMware and Virtualization ConsultantCommented:
We normally configure at 75/25 for performance.
0
 
kiwistagAuthor Commented:
So far the Write latency is down and the maximum read latency halved.

Thanks for the help above, I've decided to extend the array to RAID 5.. Now to think about the Stripe size. Currently it's 256KB Full Stripe, so I'll probably leave it at that.
0
 
andyalderCommented:
Just wondering how you got 4 drives in a BL460c, it's really a DL360 maybe? Not relevant to the problem admittedly.
0
 
kiwistagAuthor Commented:
andyalder: Sorry - you are correct, it is actually a BL480c with a P400i controller.
We have 4x BL40c's in the chassis as well.
0
 
andyalderConnect With a Mentor Commented:
Make sure you add the optional write cache battery if you use RAID 5 or performance will be poor in that case.
0
 
kiwistagAuthor Commented:
Yep - that's mandatory in our environment.
Pity some of the batteries have unpredictable lives, we've had some last several and other only showing out a year max.
0
 
kiwistagAuthor Commented:
Some good thoughts/practices mentioned which helped.
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.