ISE 2.3 P2 - GUI slowness

[csavas]

Hello,

my customer has a distributed deployment with ISE 2.3P2 - all ISE roles are running on a 3595.

2x PAN

2x MnT

8x PSN

PAN and MnT are in the same DC.

We have 4916 total endpoints and 4127 active endpoints (as of today).

We are seeing gui slowness on specific use cases:

home (dashboard) --> click on “active endpoints” --> approx. 15 sec

active endpoints --> increase value of “rows/page” to 500 --> approx. 10 sec

active endpoints (with 500 rows/page) --> select all (500) items --> approx. 15-17 sec

active endpoints --> select 100 items --> delete 100 selected items --> approx. 20 sec

I know these numbers are not so high but I was wondering if waiting up to 20sec for  a endpoint list is expected behaviour.

Customer don't want to expect this slowness as they plan to deploy round about 250.000 active sessions.

Thanks,

Cengiz

[csavas]

I forgot to add ...

PAN CPU = 2%

MnT CPU  = 3%

PAN Mem = 49%

MnT Mem  = 28%

PAN (sec) CPU = 2%

MnT (sec) CPU  = 2%

PAN (sec) Mem = 24%

MnT (sec) Mem  = 13%

PAN/MnT primary located in DC1 &

PAN/MnT secondary located in DC2  with enough bandwith and latency.

[Craig Hyps · Accepted solution]

Go to ISE Operations > Reports > Diagnostic > ISE Counters.  Verify the platform detected for each server and that it matches what you expect.  You can also get from 'show tech' at CLI of each node, but likely easier from GUI.

If VMs, verify proper resource allocations.

Consider staging ISE 2.4 node in lab and restore PAN (config) and MNT (operational) data to compare performance.  We made many updates to MnT performance in ISE 2.4 which should provide noticeable GUI response, but key that platforms detected properly.

[hslai]

Deleting endpoints consumes more CPU and IO resources and I am getting ~ 25 seconds in deleting 100 endpoints and ~70% with 2-core CPU in my own setup so it seems expected.