Resolve race condition in CP optimizer engine

We've run into what seems to be a race condition in the cp optimizer engine with multiple workers. The issue occurs when a worker finds the initial solution right around the time limit. In this case, the engine believes that it timed out, but it also believes it has a solution. This leads to an inconsistent state, that should not be possible, that causes our application logic to break. Specifically we are using the model.next() (https://www.ibm.com/support/knowledgecenter/SSSA5P_12.7.0/ilog.odms.ide.help/refjavaopl/html/ilog/cp/IloCP.html#next()) to iteratively progress our solution. next is defined to return true is a solution is found. In our case next returns true, we add some more time and an objective function, and we call next() again. However this time the call fails immediately because the engine thinks it timed out. This is not what we would expect as the original call to next returned true (meaning a solution was found). The engine should be improved to only have timed out if a solution was not found.

You can think of our pseudo code as
model.setTimeLimit(10 seconds)
if (model.next()) {
// We found an initial solution, now let's optimize
var madeProgress = true
while (timeRemaining > 0 && madeProgress) {
madeProgress = model.next() // This call fails
}
}

Post comment

By clicking the "Post Comment" or "Submit Idea" button, you are agreeing to the IBM Ideas Portal Terms of Use.
Do not place IBM confidential, company confidential, or personal information into any field.

Please enter your email address

RELATED IDEAS

Resolve race condition in CP optimizer engine