[nycbug-talk] Serving High Performance Web Sites --- UNIGROUP Meeting 22-MAY-2008 (Thu):

[George Georgalis]

On Fri 16 May 2008 at 03:39:47 PM -0400, forest mars wrote:
>On Thu, May 15, 2008 at 4:49 PM, George Georgalis <george at galis.org> wrote:
>
>> well one approach is to use squid as a reverse proxy.
>> problem with a lot of (all) alternate webservers is
>> they are not apache, so various modules and howtos just
>> don't apply to alternate. Using a reverse proxy lets
>> the proxy serve pages that don't require apache while
>> querying your *unchanged* apache should a request come
>> for dynamic or expired content. Squid is designed for
>> speed.
>>
>
>
>That's one approach for sure, though it seems like it's rate of adoption has
>dropped off, possibly bc it's seems easier to do just about anything aster
>with lighttpd. I don't know offhand any corner cases of something you
>*can't* do faster with light-y, though there may very well be some.
>
>So aside from squid adding an extra point of failure, and whether or not
>squid is designed for speed, what about the added latency to parse & proxy
>each request? Not to mention the additional hardware requirements that don't
>actually give you any real increase in capacity, even after throwing
>additional cpu cycles to compensate for the performance hit. I'm just
>saying...

a squid proxy in front of apache is much faster.

it can run on the same host, the config file is 98%
comments, probably the only doc you need, but it can be
overwhelming, at least for reverse proxy you only need
to uncomment/adjust 2 or 3 lines.

>ps to George- Hope you can make the talk on Thursday, just to keep things
>lively!

can't make it; and wouldn't want to get in a debate
anyway... works for me. :)

// George



-- 
George Georgalis, information system scientist <IXOYE><