Bill Burke wrote:
I did ran ECPerf tests and the RICE study on both linux and win2k. Linux
was 2 times slower than Win2k. Not sure why, not much of a sys-admin so
wouldn't know what to tune.
When did you try this? What versions of jboss and Java VMs on both win2k
and linux? I have no problem accepting this if you used 1.3.x from sun
on both platform, as it is very slow on linux. If this is true, can you
do the tests again using better JVMs under linux? IBM/blackdown/jrockit
1.3 on linux is much faster. 1.4 is also much faster than sun 1.3.
Some other performance tidbits:
- JBossWeb (Jetty) is 30% faster than Tomcat 4.0.4 under the heavy load
tests I did. (RICE study).
- Make sure that your Web tier and EJB tier are running in the same VM.
You'll see huge performance gains.
- Make sure you have enough DB connections in your DB pool. Otherwise your
system will start to slow down under heavy load.
- Are you having problems with Entity Bean locking? JBoss 3.0.4
(Branch_3_0) has a EntityBean monitoring MBean. In jboss-service.xml in
.../server/default/conf (or server/all/conf) uncomment out the
EntityBeanMonitor MBean. Run your stress test again, and then go to the JMX
console :8080/jmx-console and view the stats on your Entity bean concurrency
issues. If you have high concurrency and high wait times then there is a
locking problem. Consult the for-pay documentation on how to solve Entity
Bean locking problems
- Are you sure you're not overloading your DB's I/O?
- Hire JBossGroup to help out your performance. I personally have helped
out Dow Jones Indexes (www.djindexes.com) to improve their performance to
over 1000% as well as Elogex (www.elogex.com). Other JBossGroup consultants
have done the same at other sites if you need references.
- From my ECPerf tests I've found that the DB was the limiting factor.
But DB could be limiting because of jboss use it ineffective.
Bill Burke
JBossGroup
-Original Message-
From: [EMAIL PROTECTED]
[mailto:jboss-user-admin@;lists.sourceforge.net]On Behalf Of Rahul Ganjoo
Sent: Monday, October 28, 2002 8:11 AM
To: [EMAIL PROTECTED]
Subject: [JBoss-user] A question on Jboss scalability!
Hi All!
I have a few queries regarding application performance and scalability
of a web based app.
Jboss: 3.x
Tomcat:4.0.4
Op Sys: Linux(red hat 7.2),win2k,solaris
DB:Oracle 9i
H/W:Pentium 4 dual ,1GB RAM.
Now using a profiling/testing tool we saw that not more than 440 users
could use the system concurrently(on win 2000 enterprise server).The app
typically has the following archtitecture:
UI Layer: Jsps
Middleware: Session beans
DB layer: Entity beans with CMP2.0 or in some cases DAO layer(JDBC calls
from session beans)
I don't know if this info is vague or enough to tell me
1.What Jboss specific things should I take care of in order to scale up
to
1000+ concurrent users.
2.Any other tips that will enable me to scale up.
3.typically how does the performance vary from win2k to linux to
solaris.
4.Is there any upper limit on the number of users that Jboss can support
concurrently.
We are not running Jboss clusters yet.
Guys let me know if you need any other info to help me with this.
Thanks a lot in advance,
Regards,
Rahul Ganjoo.
---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
___
JBoss-user mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-user
---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
___
JBoss-user mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-user
---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
___
JBoss-user mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-user
