On Optimal Update Policies and Cluster Sizes for 2-Tier Distributed Systems

We try to analyze a generic model for 2-tier distributed systems, exploring the possibility of optimal cluster sizes from an information management perspective, such that the overall cost for updating and searching information may be minimized by adopting a judiciously lazy updating policy. We do not assume either centralized coordination or decentralization, and since it is an initial work, we only advocate the existence of such optimal policies rather than how such policies may be discovered by the system participants. We put our work in perspective using two examples from diverse domains of distributed systems, namely the wireless cellular networks, which are based on centralized coordination and peer-to-peer systems using clusters (like Kazaa).