DATA MANAGEMENT IN DISTRIBUTED SYSTEMS: A SCALABILITY TAXONOMY

Data management is a key aspe t of any distributed system. This paper surveys data management te hniques in various distributed systems, starting from Distributed Shared Memory (DSM) systems to Peer-to-Peer (P2P) systems. The entral fo us is on s alability, an important non-fun tional property of distributed systems. A s alability taxonomy of data management te hniques is presented. Detailed dis ussion of the evolution of data management te hniques in the dierent ategories as well as the state of the art is provided. As a result, several open issues are inferred in luding use of P2P te hniques in data grids and distributed mobile systems and the use of optimal data pla ement heuristi s from Content Distribution Networks (CDNs) for P2P grids. 1. Introdu tion. Data management is an important fa et of distributed systems. Data management en ompasses the ability to des ribe data, handle multiple opies (repli ation or a hing) of data obje ts or les, support for meta-data as well as data querying and a essing. Dierent approa hes for data management have given importan e to these dierent aspe ts and provide expli it support, while other aspe ts are impli itly or indire tly supported. For instan e, Distributed Shared Memory (DSM) systems and shared obje t spa es handled onsisten y of repli ated data, but supported meta-data indire tly through obje t lookups. Orthogonal to the above mentioned issues of managing data, the main non-fun tional hallenges are fault-toleran e, s alability and se urity, as illustrated in [32℄. We survey various distributed systems from the per-spe tive of s alability of data management solutions and provide a s alability taxonomy. We lassify dat