CPS Core Performance

...

Comparison of NCMP Performance across versions

Release	Date	CmHandles	100	500	1000	2000	5000	10000	20000	40000	Comments
Kohn	Oct 2022	Time	8 sec	16 sec	17 sec	33 sec	1 min	3 min	ERROR	ERROR	Error due to DB running out shared memory
London	May 2023	Time	6 sec	7 sec	12 sec	22 sec	1 min	2 min	10 min	32 min
current	Aug 2023	Time	6 sec	7 sec	10 sec	16 sec	31 sec	57 sec	71 sec	108 sec

CM-handle registration is multi-threaded, so performance may appear to scale better than linear, until the CPU cores are maxed out.

As can be seen below, CPU usage never reached 100% during the tests of the current version.

Release	Date	CmHandles	100	500	1000	2000	5000	10000	20000	40000	Comments
Kohn	Oct 2022	Time	7 sec	2 min	7 min	25 min	2.5 hour	est: 10 hour	est: 2 days	est: 7 days	Some values estimated due to time constraints
London	May 2023	Time	< 1 sec	2 sec	3 sec	5 sec	17 sec	37 sec	85 sec	ERROR	Error due to 32K limit
current	Aug 2023	Time	< 1 sec	1 sec	3 sec	4 sec	14 sec	23 sec	65 sec	2 min

Current release has exactly linear performance for CM-handle de-registration (on a single thread):