Content distribution through autonomic content and storage management

Abstract

Content Distribution has to date been addressed by a mix of centralized and uncoordinated distributed processes, such as server replication and traditional node caching mechanisms, respectively. It is an inherently distributed process that is also increasingly relying on entities that are not only increasingly distributed but also increasingly autonomous. Consequently, centralized - and typically targeting the "socially optimal© - decisions are rather unrealistic for a distributed environment of autonomic entities. Instead, a distributed management of the engaged autonomic entities, which take decisions dynamically, should be key to efficient content distribution. The latter is advocated in this paper in which two entities that are central to content distribution - specifically the content and the node storage - are considered and it is discussed how their autonomic behavior drives the operation of a content distribution network. In the first case, it is the content that manages itself by dynamically generating duplicate copies and pushing them to (seizing) the appropriate storage. In the second one, it is the node storage that is in charge, deciding on the content to be locally stored. The decisions taken by the distributed and autonomic entities may - in the extreme case - be driven by self-awareness and self-interest only, without any network state information and co-operativeness. Or, they may use (some) network information and take decisions in a more cooperative manner, despite their autonomic and self-interest-driven nature. An example is presented on the later case, showing the potential both social and individual benefits. © IFIP International Federation for Information Processing 2005.