cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Berin Loritsch <blorit...@apache.org>
Subject [C2] Running JMeter on Cocoon Samples
Date Wed, 14 Feb 2001 21:06:03 GMT
I discovered something incredible.  The XSP system is our
major performance sink-hole.  This I find to be amazing.
If you compare the /sql/sql-page to /xsp/esql then you will
find that even though the SQLTransformer is doing more
(visually) it scales much more easily than ESQL.  I then
tried running it on xsp/simple and forms/add-department
which did very little (comparably).  The forms/add-department
was placed as XSP for the sole reason that I wanted the sitemap
to be simple.

Here are the averages according to the Cocoon.log file:

Welcome  SQL     ESQL      SIMPLE     ADD
-------  ------  --------  --------   --------
96 ms    156 ms  8.64 Sec  8.29 Sec   6.68 Sec

Now, upon closer inspection, the SQLTransformer class stopped
trying to get information from the database after a certain
point.

I set JRun to simulate 1,000,000 page requests per day
(12 simultaneous threads, requesting once a second with
 a full second deviation).

To me, this deviation of performance is rediculous.  I find
XSP to be a very compelling technology, but it's real world
scalability is lacking.  Even after it has compiled, there
is too much going on in there during page generation.  The
XSP pages technically should run faster than the normal pages
because it is a class loaded, and hardcoded to send the SAX
messages right away.

I have my suspiscions as to where the problems may lie:
Class validation (is it current?) and sending too many
namespace events.  I am going to try running reading a
normal file through the LogTransformer, and then an XSP
file through the same LogTransformer.  I have a feeling
that those two areas are are major performance bottlenecks.

Mime
View raw message