A Network-aware Virtual Machine Placement approach for Data-intensive Applications in a Cloud environment

Cloud computing provides beneficial services to users, enabling them to share large amounts of information, employ Storage Nodes (SN), utilise Computing Nodes (CN) and gather knowledge for research. Virtual Machines (VMs) usually host data-intensive applications, which submit thousands of jobs that access subsets of the petabytes of data distributed over Clouds Datacentres (DCs). The VMs scheduling allocation decisions in cloud environments are based on different parameters, such as cost, resource utilisation, performance, time and resource availability. In the case of application performance, the decisions are often made on the basis of jobs being either data intensive or computation intensive. In data-intensive situations, jobs may be pushed to the data; in computation intensive situations, data may be pulled to the jobs. This kind of scheduling, in which there is no consideration of network characteristics, can lead to performance degradation in a cloud environment and may result in large processing queues and job execution delays due to site overloads. This thesis proposes a novel service framework, the network- aware VM placement approach for data- intensive applications (NADI), to address the need for improved application performance . NADI takes into account a jobs time cost based on a mechanism that maps VMs against the resources when making scheduling decisions across multiple DCs. So, it not only allocates the best available resources to a VM to minimise the time needed to complete its jobs but also checks the global state of jobs and resources so that the output of the whole cloud is maximised. The thesis begins with a statement of the problem addressed and the objectives of the research. The methodology adopted for the research is described subsequently, and the outline of the thesis is presented. This is followed by a brief introduction highlighting the current approaches in VM placement and migration in cloud computing. Next, this thesis presents a framework for the proposed NADI with a description of its various components and enabling functionalities, which are required to realise this framework. Multi-objective strategies suitable for the problems in NADI are presented. Novel algorithms for managing applications and their data are proposed; they aim to improve each jobs performance and minimise the traffic between the application and its related data. The results indicate that there are considerable performance improvements and that the completion time is reduced by 25% to 51%, which can be gained by adopting the NADI scheduling approach.