CSA 5.2 Benchmark

joe.ho · ‎10-01-2007

Can someone tell me is there any documentation regarding CSA benchmark? For example the number of sevice request (DNS) that will degrade the server performance. Is there any 3rd party web site may have those information?

tsteger1 · ‎10-02-2007

I don't know of any web sites or documentation and the test data might not apply to your situation anyway.

The overhead required by a CSA agent on a host is around 1-3%.

Other then that, I guess it depends on the server since all are different.

It also depends on the rules enforced on the agent.

The only way to get a definitive answer is to test it.

What are you trying to discover?

Tom

joe.ho · ‎10-02-2007

The client wants to see some numbers on how CSA can affect the server performance. The first one I am doing is DNS server. Other servers will be DHCP, web, Citrix and so on. If there is no such doc related to this which testing application will you recommand to run tests like this?

tsteger1 · ‎10-02-2007

Testing may be subjective because there are so many factors involved.

What app do you use now for this type of testing?

Our testing consisted of installing on test servers then moving to production servers in test mode, then in protect mode.

These are the predefined groups for servers in 5.2.225:

Servers - All types 5.2 r225 Default group for systems that install the Server agent kit Windows

Servers - Apache Web Servers 5.2 r225 Systems running the Apache web server Windows

Servers - DHCP Servers 5.2 r225 Systems running DHCP servers Windows

Servers - DNS Servers 5.2 r225 Systems running DNS servers Windows

Servers - Externally deployed 5.2 r225 Default group for servers deployed on public networks Windows

Servers - File Servers 5.2 r225 Systems running File Servers Windows

Servers - IIS Web Servers 5.2 r225 Systems running Microsoft IIS web server Windows

Servers - Internally deployed 5.2 r225 Default group for servers deployed on private networks Windows

Servers - SQL Server 5.2 r225 Systems running Microsoft SQL Server database Windows

The 1-3% overhead is pretty accurate. A rule of thumb is the more rules it has to process, the more overhead it takes.

If these are single function servers, it should work fine.

If your client wants more info I'd try and get Cisco involved.

Tom

joe.ho · ‎10-02-2007

Thanks for the detail respond. Since this request just came from the client, we are not testing anything at this point. Is there any apps that can emulate user requests on different type of protocol to perform stress test on the servers?

tsteger1 · ‎10-03-2007

I know there are web stress testers available from MS and others.

This may also be a good resource:

http://www.opensourcetesting.org/performance.php

I think you may have to pay for some of the others available.

Tom