About "mean response time"

[Dmitry Kurochkin]

Hi Erico.

On Tue, 14 Feb 2012 23:17:09 -0300, Erico Augusto Cavalcanti Guedes <eacg at cin.ufpe.br> wrote:
> Dears,
> 
> on PG docs about Run-time stats:
> (1)      (2)(3)     (4)    (5)      (6)    (7)    (8) (9)
> 000.09| i-dflt     23   4.52    196   0.00   0   1
> 000.17| i-dflt     48   5.00    199   0.00   0   1
> 000.25| i-dflt     73   5.00    199   0.00   0   1
> 000.34| i-dflt     98   5.00    198   0.00   0   1
> 000.42| i-dflt   125   5.40    185   0.00   0   1
> 
> Here is a brief explanation for each column of the run-time stats output.
> 1. Minutes since start
> 2. "i-" and "p-" stand for interval and phase-based stats
> 3. current phase name
> 4. number of replies received so far (across all phases)
> 5. reply rate (replies per second)
> 6. mean response time in milliseconds
> 7. hit ratio in percents (client side);
> 8. nothing interesting on the server side yet
> 9. number of transaction errors during that interval or phase (note: not
> all errors are transaction errors)
> 10. number of open sockets, including UDP and other housekeeping sockets if
> any;
> 11. this number is usually close to the number of pending transactions;
> (interval stats only!)
> 
> The question:
> 
> Is number 6 (mean response time) the mean time of all responses or is it
> related with number 7 (hit ratio in percents) and it is the mean response
> time of requests that were on cache?
> 

It is the mean time of all responses received by the client, including
both hits and misses.  If you are interested in detailed response time
stats for different kind of replies (hits, misses and more), they are
available in binary logs.  You can read binary logs with polygraph-lx(1)
tool or generate an HTML report with polygraph-reporter(1).

Regards,
  Dmitry

> Am I been clear?
> 
> Thanks in advance.
> 
> Erico.
> _______________________________________________
> Users mailing list
> Users at web-polygraph.org
> http://www.web-polygraph.org/mailman/listinfo/users