On 08/22/2007 08:33 PM, Vincent Bray wrote: > Hoi, > > It's a common misconception that because mem_cache uses 'memory' > rather disk_cache using 'disk', that one should assume the former is > somehow faster. > > Documenting this misconception is fraught with not wanting to say > "this sucks", so let's not say that. How about: > > Index: docs/manual/mod/mod_mem_cache.xml > =================================================================== > --- docs/manual/mod/mod_mem_cache.xml (revision 567358) > +++ docs/manual/mod/mod_mem_cache.xml (working copy) > @@ -41,6 +41,11 @@ > > <p>Content is stored in and retrieved from the cache using URI based > keys. Content with access protection is not cached.</p> > + > + <note><title>Note</title>Don't assume that because this module caches in > + memory it is faster than <module>mod_disk_cache</module>. Modern virtual > + memory systems can make delivery from disk more efficient using > kernel level > + <code>sendfile()</code> support.</note> > </summary> > <seealso><module>mod_cache</module></seealso> > <seealso><module>mod_disk_cache</module></seealso> > > (I should note that this 'patch' is an addition to the <summary> atop > mod_mem_cache.xml) > > There are of course other issue related to per-worker caches in > mem_cache, but I don't know sendfile() from sendelbow(), so this is > the best I could come up with. > > I'm not terribly diplomatic at the best of times, so is there a better > way to suggest that users try disk_cache first?
Maybe a pointer to http://httpd.apache.org/docs/2.2/en/caching.html#inmemory could help. It has some more detailed explanations. Regards Rüdiger --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
