Re: DHR is reporting low

From: Artem Veremey (averemey@digisle.net)
Date: Tue Feb 27 2001 - 19:26:15 MST


>>>> RA == Rajesh Agrawal <raagrawa@cisco.com> writes:

  RA>
  RA> 1. Is there any workload file available to test Reverse proxy.
  RA>
  RA> My setup is like this:
  RA>
  RA> [client]----[router]----(WAN)----[router]------[Server]
  RA> |
  RA> [Proxy]
  RA>
  RA> 2. Do i need to use ipfw command to introduce a delay, even if i am using
  RA> WAN setup ?
  RA>
  RA> 3. I used Polygraph 2.4 on freebsd with poly-mix3 workload to test my cache
  RA> engine (Cisco-507). I am getting perfect TPS performance. But, Hit Ratio
  RA> is only 32.53 % however, offered DHR is 57.74%.
  RA>
  RA> Following errors are reported in the report file. Is there any way to
  RA> improve DHR ?

IMHO
Desired DHR is a parameter you can specify in Polygraph config, something like:
robot R {
        ...
        recurrence = desired_DHR/cachability_ratio
        ...
}

If you are getting hit ratio 32.53% instead of something close to 55% (with is your desired DHR) it means that about 20% of the requests that should be served from cache are served from the origin server. Practically, each second request to a resource that supposed to in cache is forwarded to the server. It might happen when resource "expire" (get erased) too quickly from the cache or cache doesn't keep cachable resources.
You can improve actual DHR by improving your cache engine to fully understand and honor cache headers.
Other possibility is that polygraph doesn't work well with your cache engine and doesn't understand some of the cache responses properly. I would inspect tcpdumps of polyclt-cache-polysrv sessions together with logs.

  RA>
  RA> Errors (0.01% of all transactions):
  RA>
  RA> #errno count count% explanation
  RA> 54 168 56.38 "Connection reset by peer"
  RA> 263 130 43.62 "premature end of msg body"
  RA>
  RA> Potential problems:
  RA>
  RA> 1.Measured document hit ratio (32.53%) differs from the offered
  RA> DHR (57.74%) by -43.66%.
  RA> 2.Measured byte hit ratio (33.21%) differs from the offered BHR
  RA> (59.89%) by -44.55%.
  RA> 3.Reported fill count contribution (35.67%) does not match
  RA> `cachable miss' estimation (47.52%).
  RA> 4.Reported fill volume contribution (37.99%) does not match
  RA> `cachable miss' estimation (50.17%).
  RA>
  RA> Thanks,
  RA> Rajesh

Declaimer: I am not a polygraph expert.

-- 
Artem Veremey

Fortune cookie here: Fortune's Contribution of the Month to the Animal Rights Debate: I'll stay out of animals' way if they'll stay out of mine. "Hey you, get off my plate" -- Roger Midnight




This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:17 MDT