Looks like this was a false alarm. The test environment I was using had some random fluctuations in the results. I assumed they averaged out over multiple test runs. However it looks as if I got some bad rolls of the dice and they did not.
I have now a modified test which has <1% variation over four runs each taking 3 hours. Testing both versions, I now see negligible difference between them.
On 13 January 2017 at 10:10, Stephen Baynes <stephen.baynes@xxxxxxxxxxxxxx> wrote:
Is there a known performance fall off going 3.5.20 → 3.5.23?I am seeing a 15% to 20% performance drop on my normal download benchmark and a crude test of uploading shows a few percent slowdown.Running on a Linux derived from Debian.Thanks--Stephen Baynes
Stephen Baynes
_______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
