Optimal Caching for First-Order Query Load-Balancing in Decentralized Index Structures

Abstract

Balancing query-load in structured overlays is an important and mostly unattended problem, apart from heuristics to deal with hot-spots. We determine the optimal caching strategy - how many caches (dependent on relative frequency of queries) to reduce search latency and where to place these caches? Our query-adaptive replication scheme provides first-order balancing of query-load. We also elaborate the limitations of first-order balancing (ignored by all existing heuristics) though it should have been obvious given the experiences of the P2P community with storage-load balancing.This work lays the ground for our ongoing work on load-aware routing in structured overlays as a mechanism complementing caching to achieve query-load balancing.