[FYI] serious piggybank performance bug fixed

From: Stefano Mazzocchi <stefanom_at_mit.edu>
Date: Wed, 07 Sep 2005 02:34:53 -0400

I just committed a patch that fixes some of the performance problems in
piggybank, especially on pages that were already loaded (or were
displaying information that was loaded before).

On my laptop, the page load time went was reduced from some 15 seconds
to 500ms, which is a 30x speed improvement. All this was caused by a
runaway boolean flag that make the system lookup the statement labels in
one hashtable yet saved the results into another, resulting in continous
cache misses. It took me 5 days to figure out but I'm glad I did,
because now I understand a lot more where we spend time.

I suspect there are at least a few hundred milliseconds to reduce,
althought I would rather spend time on having faster subject-based
queries in the triple store first, which are, by far, our hotspot right now.

I also committed a change for the encoding problems we were
experiencing, which was due to the fact that our templates advertise
UTF-8 encoding but the default servlet getWriter() method returns
ISO-8869-1. Unfortunately, this still doesn't fix our problems.

I'll need to investigate further.

Stefano Mazzocchi
Research Scientist                 Digital Libraries Research Group
Massachusetts Institute of Technology            location: E25-131C
77 Massachusetts Ave                   telephone: +1 (617) 253-1096
Cambridge, MA  02139-4307              email: stefanom at mit . edu
Received on Wed Sep 07 2005 - 06:30:43 EDT

This archive was generated by hypermail 2.3.0 : Thu Aug 09 2012 - 16:39:18 EDT