A Fault-tolerance Mechanism for Future Updates 

Futures are placeholders for an awaited reply of an asynchronous method call.

The main goal of this internship is to design a mechanism that allows  the existing fault-tolerance protocol to deal with first class futures. Indeed, first class futures come with very intreresting properties that ease the programming of distributed application: whatever the order of future updates, whatever the set of processes to which this future is to be updated, the result of the computation is the same! Unfortunately, if those properties ease programming, they make the design of fault-tolerance protocols a little more difficult.

The fault-tolerance protocol that has been implemented in ProActive does not support first class futures: the value of a future can only be re-sent to the sender of the associated request upon recovery. In other words, if the future has been automatically forwarded to a third process, this process will not receive the future value upon recovery.

This internship can be addressed both from a purely theoretical approach (focusing on the link between ASP calculus and events model, see bibliography below); and from a much more applied point of view that could consist, for example in designing and implementing an independent protocol for addressing future updates.