Arbitrator-based algorithm for measurement collaboration problem

Accuracy is one of the primary goals of network measurement. Concurrent execution of measurement tasks usually impacts the accuracy of measurement results, while exclusive, sequential execution of measurement tasks may avoid this kind of influence. On the other hand, one-way measurement always requires collaboration of two nodes; thus the randomly initiated measurement tasks may produce conflicts and then lead to problems such as deadlock and low efficiency. We call this kind of problem the measurement collaboration problem (MCP), which belongs to the general distributed collaboration problem with conflicting tasks, while earlier studies mostly focus on a conflict-free solution. The MCP solution is an algorithm for deciding which measurement tasks can be performed by negotiation among nodes. According to the number of nodes with decision-making power, the solution can be divided into arbitrator-based and completely distributed. This paper gives the arbitrator-based MCP algorithm and proves the algorithm's aliveness, correctness and optimality. The algorithm can also be utilized in many aspects, such as disaster recovery, distributed-agent collaboration, object positioning and so on.