Adding a simple layer to cache the mapping of SERVER_NAME+REQUEST_URI to
cacheId should yield quite a performance bump. This would allow most requests
to bypass the entire architecture, including checking filemtimes, settings, etc.
The TTL should probably be low so that source file edits won't be ignored too
long. Even very short TTLs should yield big gains under load.
This would be easiest built into the 2.2 refactoring, but hacking it into 2.1.4
would probably be worth it.
Original issue reported on code.google.com by mrclay....@gmail.com on 5 Oct 2010 at 6:58
Original issue reported on code.google.com by
mrclay....@gmail.com
on 5 Oct 2010 at 6:58