[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: 1.7 Performance via HTTP

From: Mark Phippard <markphip_at_gmail.com>
Date: Thu, 12 May 2011 14:49:16 -0400

On Thu, May 12, 2011 at 2:37 PM, Justin Erenkrantz
<justin_at_erenkrantz.com> wrote:

> Hmm.  More specifics on your configuration would be helpful if I am to
> try to reproduce your setup.  Serf version?  (0.7.2 or higher would be
> needed to avoid the Nagle issue.)  OS (client & server)?

Serf trunk @ HEAD

Client is Windows 7
Server is Windows 2008

> You
> mentioned on the page that you are using a VPN - what type of VPN?


>> I was mainly trying to see the HTTP request count with HTTPv2 so I a
>> using a 1.7 server.
> I maintain that request count isn't an highly interesting metric -
> especially given serf's design.

Sorry, I was not clear. I was interested in the reduction of requests
from 1.6 to 1.7 when using Neon. I only ran the tests with Serf after
to compare the performance numbers. I know the request count goes
down with Serf too, but I did not bother running the tests using Serf
with 1.6.

>  Since the requests are pipelined, the
> latency impacts should be minimal - if it were serialized and
> synchronous, yes, it'd be a concern.  But, it's parallel and async, so
> that number doesn't dominate in a properly tuned environment - what
> you give up by sending more from the client, you should get back from
> having better server-side scalability/partitioning capabilities.  --

For Neon, I think it is significant. In my case, 1.7 Neon issued 6,308
fewer HTTP requests. My latency is .2 seconds. So that lopped around
21 minutes off the total test time when comparing Neon to Neon. I
realize with Serf it is less of an issue.

Mark Phippard
Received on 2011-05-12 20:49:46 CEST

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.