Performance - Changing over to Tomcat from Weblogic

[Neil Piper <ne...@adelaide.edu.au>]

Hi,

I'm finding it really hard to find performance data on Tomcat and
various other web servers in an accessible format - (conspiracy by
marketing depts?)

We run a large courseware product serving a universities online
education needs via the middleware product blackboard. In version 6
Blackboard
are considering switching from using Web logic as their supported server
to Tomcat.

We have the following statistics for a typical month:

Average successful requests per day: 330,647 
Average successful requests for pages per day: 114,235 
Average data transferred per day: 1.706 gigabytes  

At peak times we can expect a peak rate of around 100 page requests per
second on various days. (300 file requests)

The hardware in use is a 2 tier solution, an app server with Blackboard
and a database server. The app server is a SunOS 5.7 sparc
SUNW,Ultra-80. Unfortunately I don't have much performance data at the
moment on how much memory on average is in use.

Will this changeover adversely affect our performance to a large degree
? Should we upgrade hardware infrastructure (Load balancing?),  Has
anyone else tested at this magnitude of data and found tomcat to be
adequate ?   

// Neil Piper

-- 
Neil Piper

Information Technology Services
Level 7, 10 Pulteney St
Adelaide University SA 5005
neil.piper@adelaide.edu.au

--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>

Re: Performance - Changing over to Tomcat from Weblogic

[Tim Funk <fu...@joedog.org>]

I don't think tomcat can outperform weblogic, but apache with tomcat 
should easily beat weblogic. (Unless all your requests are jsp or 
servlet requests) We got rid of weblogic and went to tomcat because 
weblogic had tons of features we didn't need. And for the features we 
needed - its implementation was "sketchy". (Their interpretation of the 
Servlet API/JSP Spec and tomcat's didn't quite match. Causing us to fix 
many, many pages) Weblogic was unstable for us, had bugs which shouldn't 
have existed, and took forever to prove. Because of its instability in 
our environment, it needed restarting "fairly often" - and its startup 
time takes over a minute (compared to tomcat's almost instanteneous 
startup). Clustering/load balancing was painful because of the way the 
site was originally coded. When we switched to Tomcat - we saw some 
minor issues - which we were able to easily track down via google and in 
one case, we figured out what was going wrong via the looking at the 
source code.

Sorry for the vent - I'm sure weblogic is a fine product - but not for
what my team was doing.

That being said - tomcat with apache is hardly making a dent in our
webserver's processor usage compared to weblogic. The site I am talking
about spikes to over 100 requests a second (including images) and serves 
well over 150,000 page requests per day. We run the site on 2 webservers 
managed by a round robin loadbalancer.

But, if I were you - I would consider getting second box and put a load 
balancer in front. Apache and tomcat would reside on both boxes and if 
one hiccups/dies - the load balancer does the failover to have traffic 
only hit one box. Apache would need configured to know about 2 tomcat 
instances (one on itself - and the one on the other box) so it knows how 
to route existing sessions.

-Tim


Neil Piper wrote:
 > Hi,
 >
 > I'm finding it really hard to find performance data on Tomcat and
 > various other web servers in an accessible format - (conspiracy by
 > marketing depts?)
 >
 > We run a large courseware product serving a universities online
 > education needs via the middleware product blackboard. In version 6
 > Blackboard
 > are considering switching from using Web logic as their supported server
 > to Tomcat.
 >
 > We have the following statistics for a typical month:
 >
 > Average successful requests per day: 330,647
 > Average successful requests for pages per day: 114,235
 > Average data transferred per day: 1.706 gigabytes
 >
 > At peak times we can expect a peak rate of around 100 page requests per
 > second on various days. (300 file requests)
 >
 > The hardware in use is a 2 tier solution, an app server with Blackboard
 > and a database server. The app server is a SunOS 5.7 sparc
 > SUNW,Ultra-80. Unfortunately I don't have much performance data at the
 > moment on how much memory on average is in use.
 >
 > Will this changeover adversely affect our performance to a large degree
 > ? Should we upgrade hardware infrastructure (Load balancing?),  Has
 > anyone else tested at this magnitude of data and found tomcat to be
 > adequate ?
 >
 > // Neil Piper
 >




--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>