Re: Squid 2 - so what? ;-)

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Mon, 12 Oct 1998 22:18:12 +0200

James Zhu wrote:

> Are there any proxy server support pipelining between caches?

I said no, didn't I?

> Can you elaborate a little bit on the possible problems with
> pipelining between proxies?

Multiple problems. The major one is request interdependency. The delay
issue you mentioned is one of them. Another one are replies that force
the connection to be closed, like when the content length is unknown or
a error occurs.

> I am not sure if you meant it is hard to implement or it
> is fundamentally inefficient...

Both.

A pipelined browser connection is used by a single user. If one request
blocks/fails the connection then only that user is affected by this.

A proxy->server or proxy->proxy connection is shared by multiple users,
which makes things ugly if one users request blocks.

> Ok, so is there any other problems associated with pipelining
> between proxies?

Mostly implementation issues.

> If what I said above is the only problem, how about the idead of
> pipelining the requests from the same client through a single TCP
> connection between proxies?

That is an option, but unfortunately it does not fit to well with the
current Squid design. Not said that this design can't be changed..

> Http 1.1 clients are doing pipelining anyway, this wouldn't cause
> any extra problems. The change would be mostly modifying or adding
> a hashing function to pconn? This will make squid a lot faster,
> especially when the out going connection is slow. Pipelining over
> persistent connection gives us the real win in the performance.

True. You are welcome to give it a stab if you have the time.

---
Henrik Nordström
Spare time Squid hacker
Received on Mon Oct 12 1998 - 14:33:14 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:42:27 MST