CAS Performance Bench-Marking

[Kunal]

Can you please let me know what's the performance bench-marking is provided by CAS in terms of concurrent load CAS can support for given server configuration. I am using cas 5.0.3.

Thanks in advance!

[Waldbieser, Carl]

Kunal,

There are really a lot of variables to such a test.
I've performed load tests for CAS 5.0.x [1] and CAS 5.1.x [2], and I tried to include good summaries of my observations. Much of the load that your CAS can carry is going to depend on your deployment architechure. There is a section of the CAS 5.1.x manual that describes how to conduct your own testing [3].

Thanks,
Carl Waldbieser
ITS Identity Management
Lafayette College

[1] https://apereo.github.io/2017/06/09/cas5-perfresults-LafayetteCollege/
[2] https://apereo.github.io/2017/09/25/cas51-perfresults-LafayetteCollege/
[3] https://apereo.github.io/cas/5.1.x/planning/High-Availability-Performance-Testing.html

--
- Website: https://apereo.github.io/cas
- Gitter Chatroom: https://gitter.im/apereo/cas
- List Guidelines: https://goo.gl/1VRrw7
- Contributions: https://goo.gl/mh7qDG
---
You received this message because you are subscribed to the Google Groups "CAS Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cas-user+u...@apereo.org.
To view this discussion on the web visit https://groups.google.com/a/apereo.org/d/msgid/cas-user/fb8fe0c6-2cb3-4bca-898e-b495dc160e7a%40apereo.org.

[Kunal]

Hi Carl,

Thanks for the quick response! Appreciate the details! I have gone through this.

Actually, we ran our test with 200 concurrent login (per second) having server configuration of 4 cores and 8 GB of RAM. The average time taken by each request was 4.7 seconds, which is at much higher side, we want this time up to 1 second.

We are not able to understand, is this the idle response of CAS server OR are we missing something and what's should be the expected response time for this server configuration.

Thanks,

Kunal

[Waldbieser, Carl]

Kunal,

There could be many factors. The CAS service typically depends on some kind of back end for authentication and attribute release. And you probably have a load balancer in front of the service and are running multiple nodes. I would start by pulling my logs for the various dependant services and measuring how quickly they are receiving and responding to requests. E.g. if I am using LDAP as an attribute store and MySQL as a ticket store, I'd look at those logs and see how fast queries were coming in, and how long it took for those requests to be filled. CAS won't be able to complete certain requests without responses from those dependant pieces.

You could also try swapping out various components. E.g. do you get much better response times if you use an in-memory ticket store with a static JSON attribute store? If so, maybe tuning those components is the answer.

You didn't mention the CPU frequency you were using-- if the bottleneck is CPU-bound, that could be something to look at.

Also, it's worth mentioning, we are measuring somewhat different things. In my test, I am only interested in 6 specific authentication events (TGT create/destroy, Authentication success/fail, ST create/validate). This isn't exactly the same as performance testing.

Thanks,
Carl Waldbieser
ITS Identity Management
Lafayette College

> email to cas-user+u...@apereo.org <javascript:>.

> To view this discussion on the web visit
> https://groups.google.com/a/apereo.org/d/msgid/cas-user/fb8fe0c6-2cb3-4bca-898e-b495dc160e7a%40apereo.org.
>
>

--
- Website: https://apereo.github.io/cas
- Gitter Chatroom: https://gitter.im/apereo/cas
- List Guidelines: https://goo.gl/1VRrw7
- Contributions: https://goo.gl/mh7qDG
---
You received this message because you are subscribed to the Google Groups "CAS Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cas-user+u...@apereo.org.

To view this discussion on the web visit https://groups.google.com/a/apereo.org/d/msgid/cas-user/e615f2b3-020f-4397-bca4-9e8c9ea302d0%40apereo.org.