[midPoint] MidPoint performance & scalability on hundreds of sites

[Radovan Semancik]

Hi Tim,

The situation that you are experiencing is most likely caused by current 
implementation of midPoint GUI and provisioning. The current 
implementation is still using sequential read (GUI) and update 
(provisioning) of resources. This sequential approach provided many 
benefits in the past, mostly for troubleshooting during midPoint 
stabilization periods. But these times are over now. We have been 
planning to switch to parallel processing for some time already. The 
architecture is ready for this and the implementation should be 
relatively easy.

The root cause are especially those two issues:
https://jira.evolveum.com/browse/MID-770
https://jira.evolveum.com/browse/MID-2350

While this is obviously needed to make midPoint work well in an 
environment with thousands of resources, midPoint subscribers haven't 
chosen to "vote" for any of these issues yet. Therefore we are moving 
them out from release to release and preferring the issues that midPoint 
subscribers marked as more important. My opinion is that parallel 
provisioning in midPoint is very important. But as midPoint development 
is funded from subscriptions and sponsoring we need to listen closely to 
what subscribers and sponsors say and prioritize development according 
to that.

Therefore you have the usual options:
https://wiki.evolveum.com/display/midPoint/I+Need+New+Feature

My estimate is that if there is some way how to fund the development 
both improvements could still fit into the midPoint 3.3 release.

-- 
Radovan Semancik
Software Architect
evolveum.com



On 08/17/2015 05:03 PM, Tim.Strong at sita.aero wrote:
> Hi folks,
> Can anyone provide insight and/or real life experience on how midPoint 
> behaves & scales for literally hundreds of AD sites.
>
> We have hundreds of small AD sites scattered in all sorts of places, 
> and as I am running through my PoC to bring some test reources onto 
> midPoint, I've noticed a performance hit on midPoint when it talks to 
> a "slower site" to retrieve shadow account information.  It could be 
> slower because of bandwidth or slower because of the server hardware.
>
> While I was pulling the architecture together, I got into my head that 
> reaching out to resources/shadow accounts would be minimal since it is 
> stored in the repository.  I gather this is not the case (unless the 
> wiki is not up to date on caching, virtual approach & meta approach).
>
> Am I correct in my thinking that my performance will continue to 
> degrade the more resource accounts I link to a user?  (By performance, 
> I mean anytime a user is clicked in UI to list their info & resource 
> accounts.)  Is there anyway to overcome this?
>
> Tim
>
>
> _______________________________________________
> midPoint mailing list
> midPoint at lists.evolveum.com
> http://lists.evolveum.com/mailman/listinfo/midpoint

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.evolveum.com/pipermail/midpoint/attachments/20150818/63aa603f/attachment.htm>