A Claim-Collide Mechanism for Robust Distributed Resource Allocation

Much previous work has been done on distributed dynamic resource allocation, and there are a number of solutions. Unfortunately, pragmatic mechanisms for Internetwide resource allocation have not received much attention in the research literature. The key challenges for Internet-wide resource allocation are scale and robustness. As the number of resource managers scale, the overall resource utilization must remain high while still satisfying client requests. Furthermore, as arbitrary subsets of resource managers are partitioned from the rest, the system must still continue, with high likelihood, to allocate resources uniquely. In this paper we describe claim-collide, a CSMA/CD-like mechanism that can be used for robust Internet-wide resource allocation. Claim-collide is based on the pragmatic assumption that a network partition is healed within some specific period of time. The robust allocation however is for the expense of longer allocation latency. We use an Internetwide multicast address allocation architecture as an example of using claim-collide, and to demonstrate and evaluate through simulations how by using some carefully chosen algorithms and methods, the allocation latency can be reduced significantly. Keywords—Distributed resource allocation, Claim-collide mechanism