Mailing-List: contact dev-help@hc.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "HttpComponents Project" <dev@hc.apache.org>
Message-ID: <1477635385.121571267923147173.JavaMail.jira@brutus.apache.org>
Date: Sun, 7 Mar 2010 00:52:27 +0000 (UTC)
From: "Oleg Kalnichevski (JIRA)" <jira@apache.org>
To: dev@hc.apache.org
Subject: [jira] Resolved: (HTTPCLIENT-925) HttpClient 4.0.1 is approx. 3
 time slower then older 3.1 when works against www.google.com
In-Reply-To: <1972233046.120581267915227228.JavaMail.jira@brutus.apache.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


     [ https://issues.apache.org/jira/browse/HTTPCLIENT-925?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Oleg Kalnichevski resolved HTTPCLIENT-925.
------------------------------------------

    Resolution: Not A Problem

Oleksandr,

(1) Performance benchmarks against public web sites are utterly pointless for many reasons. For one, they cannot be reliable as most of the time you end up measuring performance of your internet link anyhow and not that of the HTTP stack.  

(2) 10 requests statistically are not enough and cannot be considered representative. You have got to be measuring performance by executing over several thousand requests over a significant time span (at least half a minute).

(3) It looks like google responds differently to requests issued by different versions of HttpClient, which also makes the results inaccurate and unrepresentative. 

I just recently changed the benchmarks to use an embedded HTTP server as a target server. You can see the updated results here:

http://wiki.apache.org/HttpComponents/HttpClient3vsHttpClient4vsHttpCore

Oleg

> HttpClient 4.0.1 is approx. 3 time slower then older 3.1 when works against www.google.com
> ------------------------------------------------------------------------------------------
>
>                 Key: HTTPCLIENT-925
>                 URL: https://issues.apache.org/jira/browse/HTTPCLIENT-925
>             Project: HttpComponents HttpClient
>          Issue Type: Improvement
>            Reporter: Oleksandr Alesinskyy
>            Priority: Minor
>
> I have tried  benchmarks from httpclient-benchmark - results are some what funny - while for most tried sites 4.0.1 is marginally faster then 3.1, for a couple of sites it is significantly slower, the most noticeable for www.google.com, test were repeated multiple times, here are sample output from 4.0.1 and 3..1  
> *4.0.1*
> Server Software:	gws
> Document URI:		http://www.google.com
> Document Length:	8459 bytes
> Time taken for tests:	3.952 seconds
> Complete requests:	10
> Failed requests:	0
> Content transferred:	84524 bytes
> Requests per second:	2.5303645 [#/sec] (mean)
> Time per request:	395.2 [ms] (mean)
> *3.1*
> Server Software:	gws
> Document URI:		http://www.google.com
> Document Length:	6847 bytes
> Time taken for tests:	1.354 seconds
> Complete requests:	10
> Failed requests:	0
> Content transferred:	68488 bytes
> Requests per second:	7.3855247 [#/sec] (mean)
> Time per request:	135.4 [ms] (mean)
> It is interesting, why.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org