Predicate matching and subscription matching in publish/subscribe systems

An important class of publish/subscribe matching algorithms work in two stages. First, predicates are matched and then matching subscriptions are derived. We observe that in practice, the domain types over which predicates are defined are often of fixed enumerable cardinality. Based on this observation we propose a table-based look-rip scheme for fast predicate evaluation that finds all matching predicates for each type with one table lookup. We compare this scheme to alternative general-purpose implementations. This observation may also suggests that matching in publish/subscribe systems could equally, well be implemented with standard database technology. We propose two DBMS-based matching algorithms and compare the better one with a special purpose publish/subscribe matching algorithm implementation. We provide first evidence that for application scenarios that require large subscription workloads and process many events a DBMS-based solution is not a feasible alternative.