Recent advances in cloud computing technology have provided companies with an opportunity to increase their profit by focusing more on their business with no worries about underlying hardware infrastructure. On the other hand, cloud operators are facing new challenges in managing and maintaining complex facility comprising a variety of computing, network, energy, and cooling infrastructure, as well as cloud software platforms.
To improve their efficiency, cloud operators require relevant information about their infrastructure, in particular the behaviour of the cloud management platform, so called the cloud middleware. A holistic analysis of cloud middleware performance will help pinpoint on its strengths and limitations. However, such analysis is very challenging due to the large number of cloud middleware components, and their siloed connections. In this thesis, we present a profiling approach based on design of experiments to meet this need.We evaluate the performance of OpenStack as a cloud middleware based on different scenarios to determine the behavior of OpenStack components and evaluate its performance.
Our experiments showed that the most influencing factor on the performance of the cloud is the number of instances created and running. Also by changing the OpenStack configurations, it can support a larger amount of workload up to 800 instances.We conclude that the performance of OpenStack is controlled by two metrics that are correlated ; the number of compute nodes and the number of workers across Nova Controller and Neutron servers. Furthermore, resource consumed by OpenStack in terms of CPU increases along with the number of instances. The current version of OpenStack (IceHouse) can not support more than 800 cases and the time required for creating instances increases linearly with the number of instances.
| Date | 4 May 2016 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Mohamed Cheriet (Supervisor) |
|---|
Abdelfattah, H. (Author),
Cheriet (Supervisor),
4 May 2016Student thesis: Master's thesis › Master in Engineering: Engineering