On Solving Some Issues in Cloud Computing

In past few years, cloud computing has emerged as one of the fastest growing segment in IT industry. It delivers infrastructure, platform, and software as a service on demand basis. Cloud provides several data centers at different geographical locations for service reliability and availability. Users can deploy applications and subscribe services from any location at competitive cost. However, this system doesn’t support mechanism and policies for dynamically coordinating load distribution among different cloud-based data centers. Further, cloud providers are unable to predict geographical distribution of users availing this services. There exist many challenging issues but few of them such as load balancing, event matching, and real-time data analysis have been addressed in the thesis. First three contributions in this thesis are dedicated to load balancing using evolutionary techniques. In the first contribution, a genetic algorithm based load balancing (LBGA) has been proposed with real value coded GA with a new encoding mechanism. Similarly, a particle swarm optimization based load balancing (LBPSO) is suggested. Both the schemes are simulated in cloud analyst, and performance comparisons are made with the competitive schemes.Consequently, both the schemes are grouped together to form a hybrid load balancing algorithm (HLBA). HLBA based central load balancer balances the load among virtual machines in cloud data center. HLBA utilizes the benefits of both genetic algorithm and particle swarm optimization. Different measures such as average response time, data center request service time, virtual machine cost, and data transfer cost are considered to evaluate the performance of the proposed algorithm. Suggested approach achieves better load balancing in large scale cloud computing environment as compared to other competitive approaches. In another contribution, an event matching algorithm has been developed for content-based event dissemination in publish/subscribe system. Proposed modified rapid match (MRM) algorithm has been compared with existing heuristics in the cloud system. Finally, a framework for the sensor-cloud environment for patient monitoring has been suggested. A prototype model has been developed for the purpose to validate the framework. This integrated system helps in monitoring, analyzing, and delivering real-time information on the fly