Re: Service Time & Response Time

Hope this helps

Thanks Sheron for sharing this!

Is this formula correct? Is this formula for calculating the wait time is confirmed by the EMC?

Hello Arasto02,

This calculation is something I use when it comes to any array Response time ~= Service Time + Wait Time

However if think you have a performance issue and wants to figure where the problem exists then there would be a lot to check to identify where the problem is.

From incorrect configuration to a failed component,  to the spindle used and the RAID chosen.

I will check if there is any documentation from EMC that talks about the same.

However if performance is what you are concerned about then you might find these links pretty useful.

http://www.emc.com/collateral/hardware/white-papers/h5773-clariion-best-practices-performance-availability-wp.pdf

Ask the Expert: Performance Calculations on Clariion/VNX

Thankyou,

Sheron

arasto02, hope the "Ask the Expert" page helped. Please let the forum know if you have any further question.

Regards,

Sheron.

Hi,

Yes, the formula you’ve mentioned is correct as per SNIA:

Wait Time = Response Time – Service Time

• Response Time is the total time from the moment an I/O request is issued until it is completed.

• Service Time represents the portion of that which the disk (or storage processor) actually spends servicing the request.

• The difference (Wait Time) reflects time spent queuing, contention delays, or waiting on resources before the disk actually begins processing the request.

Now, in CLARiiON environments through the SMI-S provider, slight discrepancies can appear depending on how the counters are polled and averaged. Tools like SolarWinds, Navisphere Analyzer, or Unisphere sometimes smooth or roll up the metrics differently, which can cause mismatches in calculated wait times.

If your Response and Service time are validated and aligned with SNIA definitions, then your computed Wait Time should also be valid. The differences you’re seeing usually arise from sampling intervals, aggregation methods, or internal cache effects.

For cross-checking, you can try shorter polling intervals and compare the deltas rather than absolute averages. That often gives a more accurate reflection of queuing behavior.

By the way, if you’re frequently doing performance calculations, tools like a time calculator, can help in quickly breaking down and validating these time differences for consistency.

Hope this clears it up!

— Alex Bhatti