number of ebase sessions per server

[Jez]

We’re running ebase server & Tomcat on a 4GB RAM, 2.66ghz CPU, Windows server.

It currently has one, small ebase app, which draws up to a max of 5 concurrent users, who are not all reading / writing data at the same time. We’ve had no hint of any performance issues to date, but are now rolling out a second ebase app which will have significantly more users – potentially up to 50 concurrent sessions. It’s a read only app and again, the users will not all be fetching data at the same time.

I’ve no previous experience of running web apps so have no idea if our ebase & Tomcat server will cope with this number of concurrent users.

Does anyone have any thoughts on number of concurrent ebase sessions and whether I may need to consider more ebase servers to cope with this level of usage – I know there are a lot of factors involved (size & complexity of app, number of reads, network, database etc etc), but I’m just trying to get a general feel.

I don’t think we will have issues at the database end as it’s Oracle 10g on an IBM P590 with stacks of memory & CPU – spec’d for up to 1000 concurrent users in an Oracle OAS environment.

[Vircos]

Hi Jez,

We have a Windows 2003 server running with 4GB RAM and a dualcore AMD Opteron CPU running for a year now. We have about 600 form-starts a day and we have not encountered any problems so far with the exception of one popular form. This form was started 720 times a day upon the normal load. We had serious problems keeping the server up and running as it was unable to cope with the total load.
We noticed that sessions were kept open to long. As a result the reserved memory for JVM became insufficient as the amount of concurrent sessions was growing. Our solution was lowering the session-time to 5 minutes and even then the load was (to) heavy for our server.

Maybe Ebase can provide us optimal server settings for different scenarios.

[Joost]

We’re running Ebase server & Tomcat on a 4GB RAM, 2.66ghz CPU, Windows server.[...]potentially up to 50 concurrent sessions. It’s a read only app and again, the users will not all be fetching data at the same time.

Should be no problem. You may do a load-/stress test using the open source software http://www.pushtotest.com/ in combination with the Selenium or TestGen4Web FireFox addon to be sure. Create a testscript (I used tg4w), fire up some testnodes, fire up TestMaker (includes one node), convert the tg4w script to a TestMaker script and on you go. There is a nice (video) tutorial on their site which explains this.
Only catch in my case was it cannot go through a NTLM/ISA proxy, so I used two systems at home which were directly (NAT) connected to the internet.

I have encountered two problems in the past. First thing to make sure is you use a "clean" businessview for the form which only includes the resources the form uses. In the past I used one BV for all forms, containing all resources. Sounded handy that time, but costs a lot of memory. Second thing is your server may contain a lot of memory, but if Java isn't instructed to use it it doesn't do you much good.

Like Vircos says, some info from Ebase on what memory/gc settings benefit Ebase would be appreciated. Besides the default settings I've added "-XX:+UseParallelGC" to my java settings. Thought it would be good, but haven't tested if it helps anything.

[Joost]

Maybe Ebase can provide us optimal server settings for different scenarios.

Maybe visit ApacheCon Europe 2009 | Mövenpick Hotel, Amsterdam | 23-27 March 2009 and visit session
* Everything Tomcat - Administering, Tuning, Troubleshooting and Developing and/or
* Performance Tuning Apache Tomcat for Production

I probably go tuesday the 24th and visit the Lucene Boot Camp as well.