Disk I/O capacity

We are experiencing some poor performance on our web application during a small stress test.  We would like to determine the IOPS consumption of the application so we can evaluate the storage system.   It is a Red Hat Linux/Oracle server running in VMware ESXi 4.1.  What tools can I use to capture the necessary data needed to analyze ?
NytroZAsked:
Who is Participating?

Improve company productivity with a Business Account.Sign Up

x
 
coolsport00Connect With a Mentor Commented:
If you are thinking it is simply disk latency the VM resides on based on disk in & of itself (i.e. a misconfigured datastore for RAID, etc.), you can SSH into the Host the VM resides on, type "esxtop" (no quotes), then press 'v' & view the DAVG statistic for the VM. If it is above 10-12ms, your underlying storage is having issues. You need either more spindles in the RAID, a different RAID, or different disk types (i.e. SAS or SSD rather than potentially SATA).

~coolsport00
0
 
NytroZAuthor Commented:
The current setup is 4 10K SAS disks in a RAID 10.  I was told this can do about 300 IOPS/sec.  Looking at the performance tab on the vSphere client it shows the IOPS around 1000 during a load.  Is this accurate?  If the disks are only capable of 300 IOPS how is it recording 1000 IOPS?  I will try the extop command.
0
Upgrade your Question Security!

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

 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
Doing the calculations, approx 400 IOPs.
0
 
NytroZAuthor Commented:
The esxtop command showed me the reads/writes per second at it verified what the performance chart showed me.  The IOPS are around 1000 and disk latency is 4ms.  this runs for a bit but eventually the database comes back with TNS listener timeout error.  Would it be safe to say that the disks are being overloaded even though latency is still low?
0
 
Duncan MeyersConnect With a Mentor Commented:
When you're preparing sizing, each 10K drive can handle 140 IOPS average. They'll peak to about 2.5 x that, so 350 IOPS per drive absolute maximum. Beyond that, response time will go through the roof
Your RAID 1/0 set should be able to produce 1400 read IOPS peak assuming your RAID array is smart enough to distribute reads across all 4 drives. For writes, it'll handle a maximum of 700 host write operations. You haven't described what the stress test is and what the I/O distribution is, but I'd say that at 1000 IOPS, you're at the limits of what 4 drives can provide so you need to add more drives. Depending on what your shared storage is, you could add more drives or add SSDs to improve performance.

Incidentally, beware of chasing benchmark numbers. What's important is the user experience, not what the load generating or benchmarking software says. If you have 10 concurrent users hitting your web app, then I'd imagine performance of your existing configuration is fine. If, on the other hand, you're getting 10,000 concurrent users, well, you're in the poo and you need more disk
0
 
Duncan MeyersCommented:
Thanks! Glad I could help.
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.