Scalable event matching for overlapping subscriptions in pub/sub systems

Content-based publish/subscribe systems allow matching the content of events with predicates in the subscriptions. However, most existing systems only allow a limited set of operators, such as comparison on primitive data types (string, integer, etc). In this paper, we consider a publish/subscribe system that supports more flexible events/subscriptions with the use of advanced, yet potentially expensive, matching operators. Examples of such operators are pattern recognizers on multimedia data and spatial operators on location data. We study a critical problem in these publish/subscribe systems, namely how to optimize the matching process for a large number of subscriptions. This is achieved by exploiting the overlap in the subscriptions and sharing the operator evaluation results whenever possible. We formulate the optimal subscription evaluation problem and show that it is NPHard. We propose an efficient d-approximation algorithm, where d is the maximum number of operators in one sub...

• Type: Conference
• Year: 2007
• Where: DEBS
• Authors: Zhen Liu, Srinivasan Parthasarathy 0002, Anand Ranganathan, Hao Yang
• Keywords: Content-based Publish/subscribe Systems, DEBS 2007, Distributed And Parallel Computing, Operators, Publish/subscribe Systems