chicken-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Chicken-users] performance issue in xml-rpc


From: Graham Fawcett
Subject: Re: [Chicken-users] performance issue in xml-rpc
Date: Tue, 5 Dec 2006 18:40:05 -0500

On 12/5/06, Daishi Kato <address@hidden> wrote:
Hi again,

Now, it's about a non-fatal performance issue.
I made the following simple (empty) example,
and tried a sample script that basically calls the RPCs several times.
I found that it takes about 40ms for each request on my machine
(Pentium4 3.2GHz),
which seems relatively slow compared to other results.
(Looks like it's less than 5ms with Apache XML-RPC.)

What do you think is the bottleneck? or is there any way to find it?
and how would you fix it?

Just a thought -- try an RPC that returns a larger message, e.g.
return a string of length >1500 bytes. Does this method outperform the
"empty" example above?

If so, you might be experiencing the "short message" problem.
Disabling the Nagle algorithm (TCP_NODELAY) or using TCP_CORK on Linux
are quick ways to "solve" this problem. Another solution is to buffer
content before sending it on the network, with the goal of sending
"packet" with their maximum payloads. I wrote to the list some time
ago, about a similar problem with http-server; many people offered
many possible solutions.

Also, if you're looking at the traffic using a packet-analyzer
(Ethereal, Wireshark, ...), look for lots of packets carrying only a
little data, as well as ~500ms gaps between packet transmissions.

Then again, it might not be a small-message problem at all. :-) Best of luck.

Graham




reply via email to

[Prev in Thread] Current Thread [Next in Thread]