Logging TC4.0 and measurement

[Roy Wilson <de...@bellatlantic.net>]

Remy,

I wondered about the impact of logging.

I'm not personally interested in comparing 3.x and 4.x, just in getting a 
database on 4.0. I also wonder if such a comparison might ignore the 
feature differences noted in the rant postings last week.
 
I am interested in what the thruput and response time numbers might mean 
to a user when the system is experiencing a certain load level and a 
certain mix of requests.

If no comparison with TC3.x is to be made, then should the AccessLogValve 
be removed from the configuration? I think it might be good to leave it 
in to see what performance people get out of the box. If AccessLogValve 
is and continues to be a hog, then maybe it shouldn't be a default? 

BTW, what did you use to do the profiling? What did you see? I'm planning 
to use OptimizeIt4.0 when I can get it and get it to attach.  

Roy

-- 
Roy Wilson
E-mail: designrw@bellatlantic.net


>>>>>>>>>>>>>>>>>> Original Message <<<<<<<<<<<<<<<<<<

On 11/8/00, 12:21:45 PM, "Remy Maucherat" <re...@apache.org> wrote regarding 
Re: TC4-m4 REquestInfoExample data, questions, modified scripts:


> TC4-m4 REquestInfoExample data, questions, modified scriptsHi,

> I did some testing myself, mainly because it's been 3 months since my 
last
> round, and Catalina has changed a bit since then.
> I also did a little bit of profiling of Catalina as a whole, and of the 
HTTP
> connector.

> When you do some testing, you may want to remove the AccessLogValve from
> your configuration, because it's incredibly slow right now. Also, by
> default, Catalina logs a lot of stuff, which is not the case with TC3.x, 
for
> example. To get a fair comparison between 2 servers, I think the level of
> logging should be similar.

> I suggest we consider not including the AccesLogValve in the default
> configuration until some optimization work is done on it.

> Remy


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

Re: Logging TC4.0 and measurement

[Remy Maucherat <re...@apache.org>]

> Remy,
>
> I wondered about the impact of logging.
>
> I'm not personally interested in comparing 3.x and 4.x, just in getting a
> database on 4.0. I also wonder if such a comparison might ignore the
> feature differences noted in the rant postings last week.
>
> I am interested in what the thruput and response time numbers might mean
> to a user when the system is experiencing a certain load level and a
> certain mix of requests.
>
> If no comparison with TC3.x is to be made, then should the AccessLogValve
> be removed from the configuration? I think it might be good to leave it
> in to see what performance people get out of the box. If AccessLogValve
> is and continues to be a hog, then maybe it shouldn't be a default?

Unless we optimize it, I think it should be left out of the default
configuration.

> BTW, what did you use to do the profiling? What did you see? I'm planning
> to use OptimizeIt4.0 when I can get it and get it to attach.

A quick run of hprof gave me enough data on what needs the most work.
The main conclusion is that the part of the code which needs the most work
is HttpProcessor.read. All the functions which make use ot it will probably
end up being rewritten too. That code is quite old anyway and it's a planned
rewrite (but always delayed because of more important issues).

Remy