You are here

Luminis 5 Performance

Submitted by squidgravy on Fri, 12/11/2015 - 17:08

We just went live with Luminis 5 about 2 weeks ago. Has anyone experienced any performance issues with the portal? We are seeing (at times) the login process taking upwards of a minute or more to complete the login process to Luminis 5. In addition to that moving from page to page can (at times) take 15 + seconds each. Interesting to note though, once the browser has cached everything in the portal we are able to move throughout the portal with ease. I have also seen that when my active session is timed out in luminis while I am still in my active browser - initiating a transaction with luminis can be very lengthy while it renegotiates a new session...

Our LP5 Topography:
Luminis LP-5.2.2.0 Patch Build 160
- Portal Tiers: 8 --> placed in pool behind an F5 load balancer (virtual nodes -vmware- 10GBs RAM 4CPUs)
- Admin Tier: 1 --> in its own pool behind the F5 (10GBs RAM 4CPUs)
- LDAP: 1 --> in its own pool behind the F5 (10GBs RAM 4CPUs)
Java Heap min / max is set to 4096 across all 10 machines...

Traffic thus far during heavy load times during the day has been about 250++ sessions per portal tier node across 8 portal tier nodes
When monitoring with Google Analytics (during peak loads) i am seeing about 275+ active users / 150+ page loads per min / 4 to 6 page loads per second.

Any help or suggestions would be greatly appreciated.

Thanks
David

Luminis Version:

Comments

That setup is almost identical to what I am putting in production very soon. Our Lum IV had 4 nodes, and each could handle ~1200-1500 sessions. I read in various places that people have seen their LP5 nodes handling ~500 sessions. Above 500, and things slow down. So for LP5, I doubled our nodes.

If your nodes are slow with ~275, something must be off in your setup, or more recent versions of LP5 are behaving differently than when people were reporting success with ~500 sessions per node.

Did you follow the tuning guide here: https://ecommunities.ellucian.com/docs/DOC-3073#jive_content_id_Performa... ? I also note that they say to set the heaps to 4096MB. You might double or triple that heap on a node or two and see if it helps or hinders.

What sort of content do you have on the initial login page/home page? Anything dynamic that might be slowing things down? Email channel? Calendar channel? Requests to external services?

(Long shot: if you are new to Luminis, the first log in on any web front always takes a long time, and each tab takes a long time. After a restart, I normally log in and open each tab one by one before putting a node back into the F5 pool.)

Did you ever resolve your performance issues? We are doing load testing now. I have noticed a bit of random sluggishness once the load gets higher than 300 per node.