Start a Conversation

Unsolved

This post is more than 5 years old

4838

October 4th, 2017 16:00

Local Storage vs. SAN Performance

I have a perplexing issue with a client SQL server. The backstory is their SQL jobs started to incrementally take longer and longer to complete. They say nothing has been done on the SQL server side. This doesn't seem to be host related, as the issue persists from host to host. As part of our troubleshooting we upgraded their hosts with 10GB NIC's to the SAN. The problem persisted. As we continued to troubleshoot, we moved their SQL VM to local storage on the host. They said that after the move the VM began performing flawlessly.

I am not really sure why the VM would be performing better on local disks. They moved from a LUN made up of 24x15K disks in a RAID10 to 2x7.2K disks in a RAID1. I completed a series of benchmark tests using FIO, HDParm, and QPerf. Every test I have ran has the SAN outperforming the local disks (in some cases by a great deal). The only thing I can really think of is latency. Their particular SAN utilizes copper instead of fiber. I can't imagine it would make this significant of an impact thought. Any ideas or thoughts are appreciated. Thank you.

Moderator

 • 

7.1K Posts

October 5th, 2017 12:00

Hello elihuj,

Based on what you have stated it sounds like there is something that is not correct in your setup. What is your Hypervisor that you are using as you didn’t state it in your post? I would check to make sure that your deployment of the storage is deployed as stated in a deployment guide.

Please let us know if you have any other questions.

2 Posts

October 10th, 2017 09:00

Hello Sam, I believe we have isolated the issue. After discussing it with a colleague, we determined that it is something higher up from the Network layer that is causing latency. As expected, the SAN did outperform local disks at all levels. The benchmark tests (netcat, especially) helped to show that without any overhead from the application, the network/SAN are performing correctly. We are now working with the client to troubleshoot their application. Thank you for the followup.

No Events found!

Top