Re: [squid-users] i'm getting a lot of output in my cache.log about status_code acl

On 12.06.2012 14:10, Eliezer Croitoru wrote:
> On 12/06/2012 04:21, Amos Jeffries wrote:
>> On 12.06.2012 11:19, Eliezer Croitoru wrote:
>>> On 11/06/2012 08:56, Amos Jeffries wrote:
>>>
>>>>
>>>> You have your debug options set to display important messages, not
>>>> just
>>>> critical ones.
>>>>
>>>> It is important to be aware your "cache deny REDIRECT" is not
>>>> working as
>>>> you designed.
>>>>
>>>> Amos
>>> .. do you have any suggestion on how to make it work to not cache a
>>> 302 code response ?
>>>
>>
>> Not with the existing design. processReply() in http.cc needs to be
>> altered to run some access directive similar to "cache" in order to
>> cope
>> with responses.
>>
>> Is this for your system with ICAP service to cache YouTube? You can
>> add
>> no-store control to the 3xx response with RESPMOD pre-cache.
>>
>> Amos
>>
> well it is indeed for this service.
> it works so great and i so proud of it becasue it the first "squid 3X
> based cache objects management of dynamic content such as youtube"
>
> the idea was somewhere in my head and i started implementing it a
> long time ago but for a more complicated setup then youtube.
> now i had the muse and took the time to actually test it all and make
> sure that it works great for every implementation i have ever put my
> hands on until now.
> i also added some code to collect live data using icap and i got only
> for youtube for the last 3 days a 600+ hits from a sum of 1900+
> requests on one server on a 10 MB DSL line.
> it's about 1.7 MB * 600 = about 9GB from 31GB sum was served from
> cache for dynamic content only.
> it's the first time that i can get statistics from squid access.log
> with just about one line of commands.
>
> the idea of no-store been done already.
> thanks for the idea.
>
> i am working on a better way to manage the objects on the cache
> itself by side to the lru\heap algorithms.
> by that i mean that by statistics on objects to shorten or extend
> their life span in the cache using htcp.
>
> i think that it's a much more efficient way to manage a cache and not
> to get into a point that you are storing thousands of files on a web
> server directory and serv them to the client.
>
> by the way there were some stories about that trying to to cancel the
> "range" requests of youtube wont work.
> well it seems like it works fine but the player as a burst\throttle
> mechanism that treats the download with specific speed to avoid
> bandwidth abuse.
> so it's better to cache the range requests so the clients will get a
> better speed from the player..

Cool. Are you able to write this up into the YouTube wiki page?

Amos
Received on Tue Jun 12 2012 - 03:14:51 MDT