Abstract
Cloud computing is gaining increasing popularity because of its higher scalability,more flexibility and ease of availability of its services. Cloud Computing enables service providers to build a large pool of resources to their customers so that users will have resources accessible on demand. To this end, cloud computing service providers use Virtualization, since it gives them the ability to effectively share resources among their users. OpenStack, a relatively new open source cloud computing platform, focuses on delivering network as a service (NaaS) using virtualization technology. OpenStack promises massively scalable cloud infrastructures. Being new, it remains to be investigated on how it delivers those abilities, and what the exact workings of its internal details are. The aim of this project is to study the internal network performance of OpenStack based on Neutron implementation. Network performance parameters like throughput, packet loss and packet delay will be evaluated under TCP and UDP traffic using IPERF benchmarking tool. This research work is partly inspired by the fact that HIOA s Faculty of Technology, Art and Design uses OpenStack for its own cloud based infrastructure. The investigation about network flow pattern confirm that VMs on the same network communicate at layer 2 and those at different networks communicate at layer 3, irrespective of their compute node locations. The results from network performance experiments showed that OpenStack Neutron guarantees a performance with virtually no network bandwidth bottleneck. In addition, the results shows that the location of machines in terms of compute node and network address affect network performance. The results also showed that OpenStack Neutron is scalable.