Unsolved
This post is more than 5 years old
31 Posts
1
4814
High average busy queue length when queue length is ok?
Hello,
I'm seeing performance information from Analyzer on the SPA of my VNX 5500 (with FAST Cache) that I have some trouble to clearly understand.
Actually the ABQL (Average Busy Queue Length) is continuously around 32, but if I check the QL (Queue Length) it is around 3 and the RT (Response Time) is around 2 ms.
Now from the Book ->
ABQL:
Desc: Average number of requests waiting at a busy system component to be serviced, including the request that is currently in service.
Comment: Since this queue length is counted only when the SP is not idle, the value indicates the frequency variation (burst frequency) of incoming requests. The higher the value, the bigger the burst and the longer the average response time at this component. In contrast to this metric, the average queue length does also include idle periods when no requests are pending. If you have 50% of the time just one outstanding request, and the other 50% the SP is idle, the average busy queue length will be 1. The average queue length however, will be ½.
QL:
Desc: The average number of requests within a certain time interval waiting to be served by the component, including the one in service.
Comment: A queue length of zero (which is average) indicates an idle system. If three requests arrive at an idle SP at the same time, only one of them can be served immediately; the other two must wait in the queue, resulting in a queue length of three.
How can I have an ABQL of 32 when I'm only having a QL of 3 especially if the RT is at 2 ms?
Thank you in advance for your help
Eric
kelleg
4.5K Posts
3
October 19th, 2012 15:00
ABQL is mesasuring the queuing when there is IO waiting. Queue Length is the average of all the IO and no IO at all times during the NAR. So the Queue Length will count busy and non-busy time into its calculation and ABQL will only count the queuing when there is IO present.
If you see a big difference between QL and ABQL, the means the data is coming in bursts - the IO is "bursty". You get a big burst of data, the Queue Length at that time is high, then you get pauses in the data and the QL at that time is zero.
Bursty IO patterns are hard to control and will lead to poor performance on the array. This is typically a host issue.
glen
Eric.Krejci
31 Posts
0
October 25th, 2012 07:00
Hi Glen,
thank you for your answer.
I have been able to track down which host is doing bursty IO. and as you said, it will be difficult to control.
but from the host perspective, I don't see any performance issue using the hosts's counters. from PowerPath or perfmon on the LUN having this high ABQL.
will check that very closely
thank you
Eric