Optimization of the structure cache for distributed data structures in overlay networks

Abstract

Overlay networks such as distributed hash tables offer a good solution to store, search and retrieve data in distributed systems. They are scalable, self-organizing and robust. On the other hand, distributed hash tables are only able to store and retrieve single data elements. This limitation is inefficient for future and sophisticated applications such as distributed social online networks. These applications require more advanced data structures like lists, sets or sorted trees in distributed systems. In previous works, we introduced an approach for Distributed Data Structures on a distributed hash table, with low overhead and delay. To improve the retrieval of this distributed data, a Structure-Cache was added, which stores further information about structures (for example a sorted tree) in advance on nodes, to allow the system to retrieve more data elements parallel and therefore be quicker. A limitation of these caches is the need to keep them up-to-date when data is added or changed in the structure. This paper, therefore, introduces different approaches to update these Structure Caches, evaluates the functionality of these mechanisms and compares them with each other.