Re: multipart/related <-> persistant
Hi Manuel,
[color=blue]
> Anyway, back to the original thread, persistent HTTP connections are
> precisely to avoid many simultaneous connections to the same server and
> the browser uses the same HTTP connection during the same page request .
>
> If you are concerned that the same user monopolizes the same connection
> for many different consecutive page requests, just reduce the keep alive
> timeout to 1 or 2 seconds and the connection will not last between
> consecutive normal page browsing.
>[/color]
Ok, when using a persistent connection, apache still as to handle all the
hits.
maybe up to 200 for just 1 page. When I have multipart/related doc apache
would have to handle just one hit and push out a block of data.
In the server multipart/related docs also have to be faster, when you can
put all
if the content in one file, it takes less IO , than gathering all info for
each user.
On a peek load, i think, it easier just to push out data in one block,
instead of
100 small blocks. Specialy when you're talking about small files ( less than
1500 mtu incl header ) On the other hand, i don't exactly know who a
browser/apache handles a KeepAlive.
[color=blue]
> Anyway, KeepAlive is not the recommended solution either. My original
> recommendation is to serve images with a separate dedicated HTTP server
> like thttpd .[/color]
Yep, an that's what we're gona use 8-)
Regards
Niels Berkers
Hi Manuel,
[color=blue]
> Anyway, back to the original thread, persistent HTTP connections are
> precisely to avoid many simultaneous connections to the same server and
> the browser uses the same HTTP connection during the same page request .
>
> If you are concerned that the same user monopolizes the same connection
> for many different consecutive page requests, just reduce the keep alive
> timeout to 1 or 2 seconds and the connection will not last between
> consecutive normal page browsing.
>[/color]
Ok, when using a persistent connection, apache still as to handle all the
hits.
maybe up to 200 for just 1 page. When I have multipart/related doc apache
would have to handle just one hit and push out a block of data.
In the server multipart/related docs also have to be faster, when you can
put all
if the content in one file, it takes less IO , than gathering all info for
each user.
On a peek load, i think, it easier just to push out data in one block,
instead of
100 small blocks. Specialy when you're talking about small files ( less than
1500 mtu incl header ) On the other hand, i don't exactly know who a
browser/apache handles a KeepAlive.
[color=blue]
> Anyway, KeepAlive is not the recommended solution either. My original
> recommendation is to serve images with a separate dedicated HTTP server
> like thttpd .[/color]
Yep, an that's what we're gona use 8-)
Regards
Niels Berkers
Comment