Locality aware distributed processes scheduling to improve e ciency using machine learning techniques

Abstract

Cloud provides a shared computing space on a pay-as-you-go model. Due to this sharing, it is di cult to execute the task e ciently in terms of time. Several factors play its parts such as process scheduling and computing requirement of other processes sharing the system. Moreover, network usage is highly depended on processes, typically processes are categorized computing and communication intensive. Such sharing of platform often degrade the performance of parallel and distributed simulations (PDS). The growth and advancements in the eld of Cloud Computing has presented its users with new challenges. Cloud Computing o ers resources to its consumers on pay-as-you-go basis. Cloud computing o ers storage, virtual machines, memory, processor and network as resources to its consumers. The more you use these resources, the costly it becomes. Parallel and distributed simulation (PDS) involves a number of logical processes (LPs) executing simultaneously while communicating with each other by interchanging small messages called packets using network. Since network is also o ered as a resource in cloud computing if a PDS involves exchanging a huge number of messages over a network, the procedure becomes expensive. In this dissertation a new way of simulating huge networks is presented that focuses on reducing network traffic.