forrest-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ross Gardler <rgard...@apache.org>
Subject Re: Locationmap caching initial cut
Date Mon, 07 Nov 2005 20:45:09 GMT
Tim Williams wrote:
> I committed the first cut at locationmap caching.  It works at the
> module level as opposed to an individual locationmap level.  

Cool stuff Tim, it's such a pleasure to create issues and see someone 
fix them :-) I *really* appreciate it (as I am sure others do).

> Since the
> improvements aren't that impressive, 

The figures you quote are reasonable, a saving of around 30-40 seconds 
on a build time of around 3 minutes is pretty good. There is only really 
the stuff for resolving skin (or views) related stuff and stylesheets 
that will benefit from this since pages are typically only requested 
once (unless in a dynamic environment). With the increased usage of such 
resources in the views work your work will, most likely, show even more 
improvement

I have a site that takes 45 minutes to build, you improvements will make 
quite a difference there (yes, it's very big, with lots of lm stuff).

Also if a site is hosted dynamically with pages requested frequently, 
this caching will improve response times for those sites (I guess).

> I'm wondering if someone can give
> me a better indication as to why the lm is slowing things up -- I'm
> not sure if that's guessing or whether there is some test to determine
> exactly where it's slowing up.

I originally created the issue as a "gut instinct" improvement. Not any 
real evaluation of what is going on. I should have made that clearer in 
the issue description, sorry.

If you turn on logging for the locationmap, you will see that either:

a) the logging statements are awful (quite possible as I threw many in 
for debugging purposes), or
b) there is lots of unnecessary duplication in the resolving of lm: URLs

Basically, there appears to be many duplicate log statements.

It's probably worth us evaluating the exact call sequence of events for 
some typical requests.

> 
> Here's what 3 runs each of the site-author did:
> 
> Without LM Caching
> ------------------
> 1 - 3:15
> 2 - 3:17
> 3 - 3:17
> 
> With LM Caching
> ------------------
> 1 - 2:48
> 2 - 2:39
> 3 - 2:39
> 
> Not exactly a news-making improvement...

Well worth the effort I think.

Thanks again.

Ross

Mime
View raw message