A Distributed Caching System in DTNs

Abstract

Many popular services in the Internet are based on a client-server architecture. This is not an optimal model in some cases. Especially, where the server becomes difficult to reach due to link or path failures or traffic overloadings. This Master's thesis presents a design, implementation and evaluation of a distributed caching system for optimization of content retrieval in Delay-Tolerant Networks (DTNs). The presented solution proposes pushing the content closer to the requesting clients, by caching it in alternative locations on the path between client and server. In this case, the clients are able to retrieve the content with lower latency and fewer network resource consumption. The approach is motivated by the decreasing price of storage in caches.

The first part of the thesis introduces the research problem on a general basis, and a literature overview is given to present the current research in the field of Web caching, proactive caching in Content Distribution Networks and caching in DTN environments. A design of the distributed caching system is then presented. Focus is to design a node that can participate in content caching. Theoretical approach is taken to motivate network topologies, routing protocols and distributions of the queries for simulations.

The experimental part of the thesis discusses the results of the simulations performed to evaluate the feasibility of the proposed solution. As a result, the presented distributed caching system is shown to increase the overall retrieval performance, as well as reduce the latency for obtaining the responses. In particular, caching policies that aggressively cache items always give the best results, however, with increasing storage costs. Moreover, the distribution of the queried items has a major impact on retrieval performance.