Browsing by Author Madumal, M.B.A. Prashan

Jump to: 0-9 A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
or enter first few letters:  
Showing results 1 to 1 of 1
Issue DateTitleAuthor(s)
2016-09-13With the widespread adoption of Internet of Things (IoT), it is estimated to become the largest device market in the world more than double the size of smart phones, PC, tablet markets combined. More and more organizations, consumers, businesses and governments are recognizing the bene ts of a connected environment of devices. This also introduce a fundamental dilemma. The data amassed from these devices will equate to an unprecedented volume. This problem propagates when data intelligence is needed, that require the collected data to be transported to a computation source, usually through a network connected to the internet. As current and forthcoming network infrastructure falls short handling this high velocity of data, the need for a robust computation model that processes data at the generated source becomes apparent. We aim to ful ll this void by proposing a novel architecture that is designed based on the Fog computation model. By leveraging the capabilities like reduced latency and data tra c inherent to the Fog computing platform, our architecture thrive to provide enhanced performance to applications that have demanding real time computation needs. Complex event processing is at the core of our computation model, simulating the environment changes captured through sensors in the form of events. Uniqueness of our architecture lies in its Fog to Cloud gateway design, which enable the computation model to take advantage of Cloud computing's resource pools. To resolve the scheduling of events to the Cloud or to the Fog node, we used a novel approach where a rule engine determines the computation source according to the resources of Fog node. To aid the rule engine, we formulated an event dependency resolving mechanism based on event trees. Also an extension of our rule engine concept is presented that predicts the resource values of the Fog node, which can be used schedule events anticipating a future system state. We explore a thorough evaluation and validation process that gauge the e ectiveness, performance and the accuracy of our novel architecture's implementation and its components. System comparison was done with a traditional complex event processing engine that has Cloud as the computation source. In our comprehensive evaluation process, results obtained con rm the implemented systems superior e ectiveness and performance in real time applications. Our study validates the need of the proposed hybrid computation model that makes use of both Fog computing and Cloud computing platforms in cases of applications with real time reaction needs.Madumal, M.B.A. Prashan