On Mon, 2012-07-02 at 09:04 +0200, Marco Gaiarin wrote: <snip> > Clearly, i can tackle some parameters (eg, the amount of data > c-icap/squid send while waiting for the full file, to prevent timeout; > or stop to ignore ''partial content'', seems now even worst that > honouring it), but i've a doubt: using a proxy, all the http connection > came form an unique internal IP; this can confuse/prevent ''fairness'' > of sfq, leading to worster performance and so, more frequent timeout? It depends on how you classify your traffic of course. If you have all your Squid connections vying for a small share of bandwidth then I guess this could be a problem. One thing that you might like to look at is the new packet marking patch for Squid (as of the later revisions of v3.2). This allows a mark to be set on the "server" side of Squid (ie. the interface receiving the data from the remote server) and for the mark to be re-set on the client side (ie. the data being sent to the client making the request). Andy -- To unsubscribe from this list: send the line "unsubscribe lartc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
