Motivated by the fact that XML is increasingly being used in distributed applications, we propose building a cooperative caching scheme for XML documents. Our scheme allows sharing cache content among a number of peers. To facilitate sharing, a distributed prefix-based index is built based on the queries whose results are cached. In the loosely-coupled sharing approach, each peer stores in its local cache results of its own queries and just publishes the associated queries to the index. In the tightly-coupled approach, each peer is assigned a specific part of the query space and stores in its local cache the results of the corresponding queries. Both approaches result in a dynamic organization of content that evolves over time based on the query load, the number of peers and the overall storage available. We present a number of associated design choices such as using a DHT for distributing the prefix-based index and a proactive cache replacement policy. We also report on a number of experiments that show the benefits of cooperative caching and highlight the pros and cons of loosely and tightly coupled cache sharing.

Abstract-On demand routing protocols that exploit local caches have received a lot of attention lately in wireless adhoc networking. In this paper, we specifically address cache management, an issue that has been a main source of criticism for the applicability of such protocols. In particular, we tackle the problem of accessing the cache content efficiently. To this end, we propose summarizing the cache content so that we achieve efficient lookups. This not only saves both the restrictive resources of the wireless devices such as computational power and energy but also improves the overall protocol performance. We use Bloom filters as summaries. Our experimental results using the ns simulator show that both resource savings and performance improvements are attained when such filters are integrated within the DSR protocol which is one the most widely used instance of an on demand protocol.