VNX 5700 performance issues..Need help

Question

Hi Experts, Need your help in recommending some troubleshooting steps for the following performance issue on VNX having response time average more than 40 ms: Details: We have VNX 5700 block only with Brocade switch fabric. The issues is ' The SQL server stops responding and hangs due to which the portal stops working. The error is IO taking longer time during the disk access. EMC is not exactly sure about the route . However they want us to clear unwanted paths between array and host to get consistence performance and to avoid SCSI and other timeout errors. On checking the host and IIS webserver the values look well within the threshold. However, looking at the SQL server , it is noticed that it gets overwhelmed with the storage response time and hence stops working. I suspect the issue of high service and turnaround time (ms) in this case and also there are lot of unwanted paths and mess up in the zoning which can be causing the issue. Query: 1. Regarding the above, I have recommendation to install the licensed version of powerpath to start with. 2. Could anyone guide on what should be the ideal response time like below 20ms to the SQL server and what threshold like 50ms would the SQL server stops responding and goes into loop? 3. Can we recommend the application specific pools containing similar speed disks so that to avoid any performance bottlenecks due to slower disks in the mixed pool? Any help or recommendations would be great from the experts. Regards

Jyothi_P_Bharat · Answer

Hi, As this is  a performance issue,I would suggest you to contact EMC Support team. Thanks Jyothi

Roger_Wu · Answer

What's the performance metrics at storage side? Is the LUN response time less than 20ms?

Anonymous User · Answer

Maximum bearable is 50 ms. But it has to be different in the case of SQL..... Thanks Rakesh

Anonymous User · Answer

Powerpath without licence wont make any difference in optimization of paths.... Second thing is look at the Skew areas if any.... will give more clarity... Thanks Rakesh

Joshua_Fawcett · Answer

This is a pretty complex question you are asking. You mention SQL server, Brocade, PowerPath, VNX...this sounds like a review of your entire environment. Ideally you should have support tickets open with EMC (for the VNX, PowerPath, and possibly Brocade - I assume the switch is supported through EMC) as well as with your host vendor, and possibly with a SQL support team for questions.

I can't tell you for sure at what point SQL will become unusable and stop responding. I would expect it would take much higher than 50 ms response times but it also depends on whether these are 50 ms sustained response times or spikes in latency of 50 ms. That may be a question that should be addressed to a SQL support person.

From the VNX array perspective, in the nar files you should be seeing response times of 20 ms or less. A spike or two above this would not be of major concern but if you are consistently over 20 ms in the nar files (at the LUN level) then you should start to look for possible performance bottlenecks (disks, as you mentioned, being one of them).

Not sure I follow your question on disks and pool recommendations. Where you put the LUNs that house your database would depend largely on what kind of performance you need on them. Enabling FAST VP on a multi-tiered pool should provide pretty good performance. But if you need response times better than that, you may want to consider using EFDs in a traditional RAID Group for a more stable performance experience.

Hopefully this helps some. Again, may be a good idea to get a case open with EMC and other vendors (if you haven't already) but if you have follow up questions, feel free to post them here.

VNX

VNX 5700 performance issues..Need help

Was this post helpful?