Abstract
Hierarchical data models (e.g. XML, Oslo) are an ideal data exchange format to
facilitate ever increasing data sharing needs among enterprises, organizations
as well as general users. However, building efficient and scalable Event
Driven Systems (EDS) for selectively disseminating such data remains largely
an unsolved problem to date. In general, an EDS has three distinct parties -
Content Publishers ({pubs}), Content Brokers ({bs}), Subscribers
({subs}) - working in a highly decoupled Publish-Subscribe (PS) model. With
a large Subscriber base having different interests and many documents
({docs}), the deficiency in existing such systems lies in the techniques
used to distribute (match/filter and forward) content from pubs to subs
through {bs}. Thus, we propose an efficient and scalable approach to
selectively distribute different subtrees of possibly large documents,
which have access control restrictions, to different $U_i$'s $in$ subs
by exploiting the hierarchical structure of those documents. A novelty of
our approach is that we map subscription routing tables in bs to efficient
tree data structures in order to perform matching and other commonly used
operations efficiently. bs form a DAG consisting of multiple trees from
pubs to {subs}. Along with our simple but adequate subscription language,
our proposed approach combines policy-driven covering and merging based
routing to dramatically reduce the load towards the root of the distribution
trees leading to a scalable system. The experimental results clearly
reinforce our claims.
Key alpha
Selective Publishing, XML, Publish-Subscribe, Routing