Framework for Scalable Dissemination Based Systems

A general framework for describing and constructing dissemination-based information systems (DBIS) is described in [14]. By combining various data delivery techniques, the most efficient use available server, and communication resources, the scalability and performance of dissemination-oriented applications are enhanced.

The approach distinguishes between three types of nodes: (1) data sources provide base data for application, (2) clients consume this information, and (3) information brokers add value to information and redistributes it. Information brokers bind the different modes of data delivery and drive the scheduling to select a particular mode, depending on its access patterns. Brokers provide network transparency to the clients. Brokers can also be the data sources. Data can be cached at any of the many points along the data path from the server to the client. Cache invalidations and refresh messages need to be sent to each client cache manager. LRU or some other cache replacement policy can be used in this approach. Intermediate nodes can simply pass/propagate the data or can also perform some computations over those data. Provisions are also kept to recover some nodes from failure. The server relies on the clients' profile to optimize the push schedule. The framework provides techniques for delivering data in wide-area network settings in which nodes and links reflect extreme variation in their operating parameters. By adjusting the data delivery mechanism to match these characteristics, high performance and scalability can be achieved. The toolkit provides a set of classes that allow distributed nodes to negotiate and establish a connection and local cache. The data transmission policies need to be agreed upon between the server and the clients.

Was this article helpful?

0 0

Post a comment