MP-Index: A Multi-predicate Publish/Subscribe Mechanism for Internet of Things

The IoT (Internet of Things) is a network ofphysical devices (i.e. objects like Vehicles, Machines, Buildings, Furniture, and so on) that helps achieve better service andvalue. Large amounts of data is generated while connectingthese devices and communicating between them. Faster capturing, processing and forwarding of this data certainly requiresan efficient technique during each stage of communication. The Publish/Subscribe communication paradigm is a powerfulmedium for IoT devices to connect and communicate with eachother. It offers anonymous, many-to-many and asynchronouscommunication which plays a very significant role in designinghighly scalable distributed systems. Event matching is oneof the key aspects of Publish/Subscribe paradigm where anevent that has occurred is matched against the large volumeof subscriptions to find the appropriate subscribers that needto be notified. The existing event matching solutions are based on the assumptionthat an assignment (group of events) will be generatedand matched against a large number of subscriptions. In thereal world, events do not get generated together and henceconsidering them as an assignment is not the best way to handlethem. We are proposing a solution based on the temporal, causal and spatial properties of events, which overcomes theassumption of assignment generation. Our proposed solution, MP-Index (Multi-Predicate Index) allows the Publish/Subscribesystem to match individual events occurring at different timestamps against a large volume of subscriptions. The proposedsolution was implemented and its results were compared withtwo of the existing algorithms. The results of the comparisonshow that the proposed Multi-Predicate Index algorithmoutperforms the other two algorithms in terms of matchingtime. We have also implemented the proposed solution inMosquitto which is an open source implementation of theMQTT (Message Queuing Telemetry Transport) protocol. Theresult of this implementation clearly confirms the support formulti-predicate subscription match against events occurring atdifferent time stamps.