Kumar, Prashant creator Dailey, Mathew N. Chairperson Duboz, Raphael Examination Committee Vilas Wuwongse Examination Committee Asian Institute of Technology Fellowship Scholarship Donor text series technical report th Pathum Thani, Thailand Asian Institute of Technology 2012 continuing eng 1 online resource (65 p.) : ill. Managing multi-tier applications running on a cloud is more difficult than managinga single tier application, as the requests in a single tier application get processed byonly one tier. In the multi-tier case, a request is processed in a pipeline through morethan one tier. Bottlenecks in one tier can cause increases in the response time andcan violate SLAs. Hosting multi-tier Web applications introduces many challenges forcloud providers to deal with, for example provisioning of resources, detecting workloadchanges, load balancing among tiers, optimizing tier placements, reducing migrationcosts, and increasing resource utilization in data centers. To meet these challenges,cloud providers have to develop algorithms and policies to manage their resources andrun applications efficiently.These algorithms and policies need to be evaluated and tested before they are imple-mented in the data center. Running evaluations and tests on real clouds is not feasiblegiven time and cost constraints. An alternative is to run the evaluation on a computerbased simulator that is fast, cheap, and produces nearly same result as when the test isperformed on a real cloud. Reconfiguration overhead and the cost involved in experi-menting with a model for a system is high and a simulator can provide a more realisticand detailed view of the system than the model itself. One of the main objectives ofthis research focuses on the design of a simulator that is capable of simulating multi-tierWeb applications on cloud infrastructure.Another major issue in cloud computing is to find the optimal placement plan formapping virtual machines to physical machines. This is one of the important policiesin a data center for maximizing the resource utilization and minimizing the activenumber of physical machines.In this research, we propose a placement algorithm thattries to minimize a stochastic objective function in order to find the best approximatesolution for the placement. A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Computer Science, School of Engineering and Technology Thesis (M.Eng.) - Asian Institute of Technology, 2012 Cloud computing Asian Institute of Technology. http://203.159.5.9/ait-thesis/detail.php?q=B02115 http://203.159.5.9/ait-thesis/detail.php?q=B02115 161122 20260817172945.0