On 11/05/2013 06:34 AM, Mark DeCheser wrote:
> I'm cooking up a custom log in Squid, and one of my planned data points is
> Squid response time (tr), as defined here:
>
> http://www.squid-cache.org/Doc/config/logformat/
>
> I am not running a caching proxy.
>
> Question 1:
>
> Does the response time value reflect the time spent transferring a file
> from the remote web server to the proxy server, or the time spent by the
> squid daemon handling the request in total?
Both. Often, especially when dealing with large responses, receiving
content (from the server) and sending the same content (to the client)
overlap in time. Roughly speaking, %tr timer starts when Squid starts
receiving the request from the client and stops when Squid stops sending
the response to the client.
Please note that transaction response time may depend, in part, on
"external" factors such as origin server response time and DNS server
response time.
> Question 2:
>
> Can someone demonstrate the usefulness of capturing this value for the
> purposes of tuning or troubleshooting the Squid proxy?
Well, response time is a fundamental performance characteristic
affecting both internal proxy concurrency levels and user-perceived
"responsiveness" of the network. Without response time, performance
results are usually meaningless.
For example, a proxy dealing with faster origin servers will usually be
able to sustain higher loads (due to lower transaction concurrency levels).
%tr is included in the default access log format.
HTH,
Alex.
Received on Tue Nov 05 2013 - 15:37:12 MST
This archive was generated by hypermail 2.2.0 : Wed Nov 06 2013 - 12:00:08 MST