RE: Apache Module (Xalan?) - XSLT performance study?

That is completely different:
 -A round trip trough a HTTP server can amount to nearly nothing if the
  server is not very busy and the network connection is fast;
- But a database connection depends A LOT on the operation and on the
  database itself.

A database parentesis:
(
  Example on database dependency: I benchmarked in Oracle8 vs. MS SQL
  Server 7 databases 2 simple queries. One returned a very small dataset
  and the other a very large one.

  There was not much difference in times for the large one but there was
  always a larger latency for the small one. I had the MS SQL Server 7
  in the local machine and the Oracle 8 server in another PC in the
  network but I was getting a 0 ms time for Oracle (mening that the time
  was so small that I could not measure it) and 30 ms for SQL Server.
  And yes, I had TCP/IP in SQL Server, not named pipes (which are slower).

  I admit that my tests were not very scientific but these times were
  very consistent across several repetitions.
)


I suggest we agree on some specific benchmark we can perform in both
parsers. I have an eye on Xalan and I am interested to move to Java
as soon as I can. So, I am interested also on Java parser timmings.

I will help as I can performing such a benchmark. I thing my
experience with MSXML is already good enough for me to be of some use.

We should obviously let databases and HTTP roundtrips out of the
equation.

I would like to have suggestions on how this can be accomplished, or
just some data if someone already did it.
=:o)


Thanks and have fun,

Paulo


-----Original Message-----
From: owner-xsl-list@xxxxxxxxxxxxxxxx
[mailto:owner-xsl-list@xxxxxxxxxxxxxxxx]On Behalf Of Jon Smirl
Sent: Friday, June 02, 2000 4:08 AM
To: xsl-list@xxxxxxxxxxxxxxxx
Subject: Re: Apache Module (Xalan?)


To: <xsl-list@xxxxxxxxxxxxxxxx>
> MSXML3 sure is much faster than that!!!

Have you benchmarked MSXML3 versus the C version of Xalan? My times are for
a complete round trip though an HTTP server including the page transform.
XSL is just part of the equation, there's a SQL server involved too. I'm
also still undecided about the impact of having three processes (Apache, app
server (JServ/XT/XP), dbengine) vs a single process scheme.

Has anyone done any pure benchmarks of just a cached XSLT transform on
MSXML3 vs Xalan?

Jon Smirl
jonsmirl@xxxxxxxxxxxx



 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list


 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list