CSR-DMSS, SM: Design and Evaluation of a Scalable Meta-Event Dissemination System

点击翻译按钮获取中文摘要

CSR-DMSS, SM: Design and Evaluation of a Scalable Meta-EventDissemination SystemProject Abstract:Event dissemination systems offer a powerful middleware fordistributed applications such as electronic trading platforms, newsaggregators, and network intrusion detection systems. Such systems areusually based on the publish-subscribe paradigm: published events aredelivered to clients who have a supplied a matchingsubscription. While most current publish-subscribe systems supportsubscriptions that evaluate to true/false on a per-event basis,emerging applications need support for subscriptions to higher level"meta-events", which correspond to combinations of basic events. Thechallenge in tracking such subscriptions is that there is no singleevent, when processed in isolation, permits the evaluation of asubscription to true or false.This project designs, implements and evaluates an event disseminationsystem that supports tracking of meta-events. The system convertsuser-defined subscriptions into "triggers", and efficiently maintainsthe state of the triggers using algorithms inspired by the literatureon data stream processing. The system scales with increasing numbersof subscriptions through merging triggers by identifying commonalitiesbetween them, and through indexing triggers for fast detection ofmatching subscriptions for events. The system is evaluated using aninternet-scale testbed, such as Planetlab. Such a meta-eventdissemination system will provide a more expressive interface to thedesigner of distributed and web applications, thus transforming theway in which distributed applications are authored. This is expectedto impact the application of information technology in various fieldssuch as healthcare, finance, and security. The project enhancesundergraduate and graduate education through strengthening thefreshmen engineering course in Electrical and Computer Engineering andthrough developing a strong algorithmic foundation for researchstudents in networking and distributed computing.