> ons 2010-09-15 klockan 14:01 +0100 skrev Martin Sperl: > > Essentially we see that during peak hours the Average response time is > better than during off-peak hours. > > Average response time measured on what? Measured via the access log of squid. > > Every request being handled, or the response time of some well known > synthetic request? All requests show this statistical behavior Also our "synthetic tests" that are measured externally every 5 minutes. > > If measured over all requests then it may well be normal. Under low > traffic conditions a couple slow requests such as downloads to a low > bandwidth client has big impact on the statistics, while under high > traffic those drowns in the rest of the traffic. > > I am afraid you need to drill down a bit in the data to tell what this > really is about. May be perfectly normal or may be a sign of problems. > Can't tell from the statistics alone. I know, that is why I have created and added the histogram data, that shows that say at 6am UTC >90% of the requests are above 0.03 seconds. While during peak hours (say 6pm) we have a peak at 0.011 seconds and only <5% above 0.030 seconds As Amos has said: if there are not enough "requests", the event driven design may be the "culprit" for the introduced latencies... I will need to investigate this... Thanks, Martin This message and the information contained herein is proprietary and confidential and subject to the Amdocs policy statement, you may review at http://www.amdocs.com/email_disclaimer.asp
