With the AF SDK, I know that it is possible to specify which PI Collective Member is being read from using PICollectiveMember instances.
However, I am using the AF SDK in some ACE calculations and noticed that in a single ACE executable, if I have one context connect to Collective Member A, and another connect to Collective Member B, the contexts do not appear to maintain separate connections to the distinct collective members - it seems that the collective member connected to is the last one specified across the entire client (in this case, the PIACEClassLibraryHost.exe process running my ACE calculation, and as a result, all running contexts end up using the same collective member).
Is there anyway to have two different threads from the same process connect to two different PI Collective Members simultaneously? I would like to do this for load balancing rather than high-availability purposes.
If this should be possible, what might I be doing wrong?