W3C home > Mailing lists > Public > ietf-http-wg@w3.org > July to September 2014

Re: proxies are dangerous slide

From: Willy Tarreau <w@1wt.eu>
Date: Mon, 21 Jul 2014 21:52:55 +0200
To: "Cullen Jennings (fluffy)" <fluffy@cisco.com>
Cc: Mark Nottingham <mnot@mnot.net>, "ietf-http-wg@w3.org" <ietf-http-wg@w3.org>
Message-ID: <20140721195255.GD28569@1wt.eu>
On Mon, Jul 21, 2014 at 07:28:15PM +0000, Cullen Jennings (fluffy) wrote:
> 	
> I?d like to add ?
> 
> Proxies always add latency in some cases because of the speed of light and the triangle inequality 

That's clearly far from often being true.

First, it's very common that proxies are located on fast networks, at
efficient locations close to the clients and with fast enough access
to the origin servers, while direct connections can go through the "junk"
route on links which are already congested by P2P traffic and other things.
That's for the latency alone. There are even some CDN companies who make
their business reducing end-to-end latency using carefully placed proxies.

Second, concerning a page load time, it tends to be the opposite in fact,
proxies can often *reduce* the page load time precisely because of the
speed of light : when you have too small a TCP window between a client
and a server, having a proxy in the middle cuts the RTT by two and can
double the data rate between them, simply by acking the server faster
and receiving ACKs faster from the client. In fact, the proxy serves as
an extra window over the wire.

So that's not *that* black and white. Having a poorly placed proxy will
often make your experience worse, but a smartly placed one can improve
it.

Regards,
Willy
Received on Monday, 21 July 2014 19:53:39 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 30 March 2016 09:57:09 UTC