Insufficient memory error during clustering

[Bernd Welter]

One of my customers faced the following error message while perfoming clustering:

{
"faultInfo": {
"$type": "ProcessingFault"
},
"message": "Solver problem: Out of memory"
}

Turns out that his clustering scenario was based on a HUGE number of com.ptvgroup.xserver.xcluster.PlanningLocation objects (>100'000!).

Now though the error message could be understood as "gimme more hardware/memory to get rid of the error" I'd recommend to evaluate a changed strategy in the clustering:

• Reduce the planning complexity by applying a pregrouping of locations, e.g. based on post code areas or street names

Best regards,
Bernd

[Prakash]

Hi Bernd,

I have discovered this link (https://support.gurobi.com/hc/en-us/art ... -condition) while searching solution on the internet about this issue.
Those suggestions mentioned (1-5), are we able to change the settings for communication between xCluster and the Solver for testing?

Regards,
Prakash

[ahn]

Hi Prakash,

Thanks for your question. We chose the values of xServer and gurobi parameters carefully to be able to efficiently solve a variety of instances. So the settings (also those specified by gurobi) cannot be changed.

But as mentioned in gurobi thread, the fundamental problem seems to be insufficient memory, and the best solution would be to increase memory.

Bests
Nitin Ahuja

[Bernd Welter]

Just some fancy experiment... I took almost 10'000 locations with zipcodes and activities.

• 

  the native 9288 individual locations

• 

  preaggregated into 2779 representing "zips" with accumulated activities

From my perspective you wouldn't loose quality if you aggregate...
But you need less resources.