Kompleks olay işleme kullanarak IoT uygulamalarında veri trafiği yönetimi

With the proliferation of heterogeneous services in IoT, the cloud of things has begun to face latency-based network challenges. As a solution, the fog of things (FoT) paradigm has emerged. FoT promises numerous advantages for real-time IoT services by offloading computation from the cloud to resource-constrained edge networks. Therefore, to utilize these advantages, an effective data traffic management and resource provisioning approach is needed. This thesis addresses the following three problems associated with IoT data traffic management in a scalable FoT system from the perspective of latency-sensitive applications. Problem-1: an efficient data traffic scheduling mechanism is needed to support multiple services having different QoS characteristics where the decision outputs of these services are transmitted with multi-priority traffic patterns. Problem-2: transmission delay faced on reliable data transfer between fog components should be reduced to experience better QoS. Problem-3: a sensor data freshness model is needed to understand and achieve stable integration of FoT services. To this end, the contributions of the thesis are threefold; First, a CEP-based scheduling policy is proposed for the mitigation of the starvation observed in multi-level queue design. In the proposed approach, arrival-service model of multi-priority FoT data traffic is given using finite-size multi-level waiting lines and then arrival-service waiting parameters are derived. The proposed scheduling policy is compared with FIFO and MPDQ policies through a comprehensive analysis of the waiting times and wait-time gap characteristics. Second, a message queue-based data transfer method is studied to reduce round trip delays between the components of the fog system. A comprehensive analysis of the queue-based messaging model and the conventional subscribe-update-notify-based data flow approach is given in terms of timing constraints and delay characteristics. In the analysis, the factors that result in data transmission latency are examined by formulating end-to-end latency faced in sensor and decision data for both methods. Third, a stateful freshness model is devised to quantify the sensor data freshness rate required by the application. An analysis for the behavior of the overall system in the context of data freshness is given and the proposed method is compared with the state-of-the-art stateless data freshness time calculation methods. Overall, comprehensive tests are conducted in simulations and on real a testbed environment as a proof of concept for the proposed solutions. The results reveal that the proposed solutions provide significant contributions to the FoT design in the given setup. It is also shown that, the proposed scheduling policy reduces the waiting time gaps and latency for a fog system scaling up to 800 clients and 256 sensor devices.