Cloud network performance analysis : an openstack case study

The last decade has witnessed the fast growth of Cloud Computing (CC) paradigm in the ICT world, drawing lots of attention from academia and industry. The increasing popularity of cloud operating systems, supported by the vastly decreased cost of commodity hardware, makes deploying and managing a CC data centre more feasible than ever. So far, OpenStack has been the main character behind some of the most successful stories in the Cloud service market nowadays. Academic literature, however, has little insight in how OpenStack could empower raw hardware infrastructure and turn it into scalable, on-demand pools of computing resources that can satisfy even the most computationally intensive applications and services. This thesis work aims at deploying a fully functional OpenStack cluster with different settings and operational environments, followed by an in-depth analysis of its network performance and explanation of what happens behind the scene. Using a methodological approach and having carried out numerous experiments, we present various scenarios where OpenStack Virtual Machines perform at their best and worst with regards to network performance. Eventually we are able to draw conclusions on the impacts that the Networking module places on the overall OpenStack network performance.