Re: [squid-users] squidclient mgr:info squid performance

From: Amos Jeffries <squid3_at_treenet.co.nz>
Date: Fri, 16 Dec 2011 01:35:21 +1300

On 16/12/2011 12:51 a.m., Benjamin wrote:
> Hi,
>
> what is the meaning for "Select loop called: 730321522 times, 0.117
> ms avg"
>
> squidclient mgr:info
>
> Connection information for squid:
> Number of clients accessing cache: 0
> Number of HTTP requests received: 32626996
> Number of ICP messages received: 0
> Number of ICP messages sent: 0
> Number of queued ICP replies: 0
> Number of HTCP messages received: 0
> Number of HTCP messages sent: 0
> Request failure ratio: 0.00
> Average HTTP requests per minute since start: 22877.7
> Average ICP messages per minute since start: 0.0
> Select loop called: 730321522 times, 0.117 ms avg <-----------------
>
> Is it showing that , squid does in/out for objects with this
> (730321522) times. is it right perception?

Squid has performed 730321522 tests to check for I/O events, an average
of every 0.117ms. That is all.
Several read/write/connect/close/DNS handling events may have happened
on each of those checks.

>
> My full output of squidclient mgr:info. Please tell me that my squid
> is performing well or do i need to do more tuning or something?
>
>
>
> squidclient mgr:info
> HTTP/1.0 200 OK
> Server: squid
> Mime-Version: 1.0
> Date: Thu, 15 Dec 2011 11:42:25 GMT
> Content-Type: text/plain
> Expires: Thu, 15 Dec 2011 11:42:25 GMT
> Last-Modified: Thu, 15 Dec 2011 11:42:25 GMT
> X-Cache: MISS from CACHE-Engine
> X-Cache-Lookup: MISS from CACHE-Engine:3128
> Connection: close
>
> Squid Object Cache: Version 3.1.10
> Start Time: Wed, 14 Dec 2011 11:56:16 GMT
> Current Time: Thu, 15 Dec 2011 11:42:25 GMT
> Connection information for squid:
> Number of clients accessing cache: 0
> Number of HTTP requests received: 32626996
> Number of ICP messages received: 0
> Number of ICP messages sent: 0
> Number of queued ICP replies: 0
> Number of HTCP messages received: 0
> Number of HTCP messages sent: 0
> Request failure ratio: 0.00
> Average HTTP requests per minute since start: 22877.7

Okay. 381rps aint too shabby. Highest-performance should be able to
reach at least 950rps with a lot of fine tweaking and tuning, but of
course that depends on the clients actually being there making request
and relatively fast themselves.

> Average ICP messages per minute since start: 0.0
> Select loop called: 730321522 times, 0.117 ms avg
> Cache information for squid:
> Hits as % of all requests: 5min: 33.2%, 60min: 32.1%

Fairly good for a forward proxy. You might be able to tweak it up into
the 40+% range. But that depends on the clients traffic to determine
what you can do and what will work.

> Hits as % of bytes sent: 5min: 21.0%, 60min: 17.7%
> Memory hits as % of hit requests: 5min: 16.0%, 60min: 16.0%

Memory hits is where you get the most speed. You might want to see what
is causing this to be low and see if you can tune it up a bit. But
again, that depends on the clients traffic.

> Disk hits as % of hit requests: 5min: 64.4%, 60min: 64.1%
> Storage Swap size: 304802412 KB
> Storage Swap capacity: 77.7% used, 22.3% free
> Storage Mem size: 2535920 KB
> Storage Mem capacity: 100.0% used, 0.0% free

Um. 2.5GB of RAM caching going on and ony giving you 16% hit ratio. This
does not seem great.

> Mean Object Size: 28.43 KB
> Requests given to unlinkd: 0
> Median Service Times (seconds) 5 min 60 min:
> HTTP Requests (All): 0.24524 0.25890
> Cache Misses: 0.35832 0.35832
> Cache Hits: 0.00919 0.00919
> Near Hits: 0.15048 0.15048
> Not-Modified Replies: 0.00000 0.00000
> DNS Lookups: 0.13638 0.13638

DNS appears to be slow. ~130ms. I would hope for at least one more zero
after the decimal point, making it under 100ms on average. But this may
be limited by the amount of time already spent processing requests, so
"it depends".

Amos
Received on Thu Dec 15 2011 - 12:35:30 MST

This archive was generated by hypermail 2.2.0 : Thu Dec 15 2011 - 12:00:03 MST