Voice-over-IP (VoIP) has become an important service in the Internet. In contrast to the Public Switched Telephone Network where the delivery of all messages and streams is the responsibility of the calling parties' providers, VoIP media data is sent directly between the user agents without provider interaction in most cases. Hence, a VoIP provider is not aware of media connectivity, i.e., whether a call was successful or not. This may lead to incorrect behavior when a VoIP provider offers services beyond signaling (for example, SPIT prevention, payment). In this paper, we discuss several approaches for the detection of media connectivity and present a solution that conforms with the existing standards. The modified behavior of the user agents, the use of SCTP and provider's awareness of media connectivity are described in detail. Finally, measurements show that our solution results in neglectable overhead.

We propose an approach to distributed Answer Set Solving based on Message Passing. Our approach aims at taking advantage of modern ASP solvers rather than proposing a genuine yet involved parallel ASP solver. To this end, we rely upon a simple master-worker architecture in which each worker amounts to an off-the-shelf ASP solver augmented with a separate communication module being only lightly connected to the actual solver. The overall communication is driven by the workers' communication modules, which asynchronously exchange messages with the master. We have implemented our approach and report upon an empirical study demonstrating its computational impact.

Loosely coupled master-worker applications are very well suited for Grids. In this paper, we describe our experiences with the development of a resilient master-worker application using MPI. It is well-known that the MPI standard does not address fault tolerance sufficiently, leaving de- tails mainly to implementations. Our experiments demonstrate that the most popular implementations, MPICH2 and OpenMPI, do not handle faults grace- fully. Further, we discuss different approaches for cir- cumventing these limitations in relation to their appli- cability in Grids.