Re: [squid-users] Quick statistic that shows performance improvement?

Hi,
How do I measure the non-proxied service time? (Sorry if the question
sounds stupid. I appreciate if you lead me to where to find out.)

On Sun, Mar 27, 2011 at 7:02 PM, Amos Jeffries <squid3_at_treenet.co.nz> wrote:
> On Sun, 27 Mar 2011 06:33:18 -0400, January Sharp wrote:
>>
>> Hi,
>>
>> We installed Squid 3.1.10 three days ago, and lots of clients are
>> using it.  Being a complete newbie to this, may I ask a question.
>>
>> What is best statistic to use to find out the general, overall
>> improvement of our clients' web surfing with vs. without squid.  That
>> is, when I report to my management, "Our users are able to surf X%
>> faster since we installed squid".
>>
>> Is this X available or computable from the information in the General
>> Runtime Information page of the Cache Mgr? Or somewhere else in Cache
>> Mgr?
>>
>> I am guessing it is 1 - (Cache Misses Median Service Times / HTTP
>> Requests (All) Median Service Times).  Is this correct?
>
> Any of the metrics which provide both HIT and MISS details can be used as
> performance gain measures.
>
> Speed improvements are complex. For absolute statements like that you will
> have to combine the HIT service time with a non-proxied service time.
> Mediated in some calculation against the request HIT ratio.
>  The results are rarely what you expect. Adding a non-caching proxy usually
> results in speed LOSS from the overheads, but not always.
>
> The other metric used are Byte Hit ratios/totals. Combined with your
> upstream bandwidth costs that measures a direct $$ figure in the budget.
>
> Amos
>
>
Received on Sun Mar 27 2011 - 23:33:05 MDT