Sammendrag
The necessity of allowing time-sharing a single computer between multiple, single tasking operating systems emerged the concept of virtualization. As a result, old concept of one-operating system for one-server paradigm is rundown. Of course, proper division of resources in a single machine for multiple operating systems is a tremendous job of virtualization, and hence it is flourishing among computer users. On the other hand, it is making the users confused to select the best from the pool of good operating systems, hypervisors and other tools and techniques. Similarly, choosing the number of virtual machine is not only a complex issue for them, but also the performance of the whole environment may go down on wrong judgement. Rather than doing scientific analysis on these matters, usually decisions are made on general discussion and perception.
Therefore, it is immensely desired to understand the performance impact on a virtual machine on the presence of its neighbours. This thesis scientifically evaluates and figures out the scalability impact on a virtual machine. The chosen operating system is one of the most popular operating systems, i.e., Debian GNU/Linux OS. Similarly, KVM (Kernel-based Virtual Machine) is chosen as a virtual machine monitor. The performance of the virtual machine is analysed by running series of benchmarks one by one. First the system is measured as a white box for analysing its micro-performance, and then it is measured as a black box for macro-performance. The finding of this research not only delivers an ideal pattern of system performance, but also gives a clear vision to the system administrators and/or organizations to understand how the scalability of machines impacts performance of the system, which component is less affected and which is affected more.