faculty: "FNWI" and publication year: "2011"
| Author||Gijs van Lammeren|
|Title||Automatic Scaling of the Windows Azure Platform : A study on the efficiency of scaling policies|
|Supervisors||Tijmen van de Kamp, Jan van Eijck|
|Faculty||Faculty of Science|
|Programme||FNWI MSc Software Engineering|
|Abstract||Cloud Computing is maturing and more companies are examining the opportunities that it brings to their business. Due to the elastic characteristic of cloud environments it became possible to adapt the recourse usage to the workload. This minimizes the waste of idle resources and because of the pay-as-you-go model the costs are more
manageable. The process of dynamic resource provisioning can be automated, but only a few cloud vendors offer this out-of-the-box. The reason is that different applications have different scaling needs.
This research focusses on the automatic scaling of web applications hosted on the Windows Azure Platform. Application owners aim to scale as efficiently as possible: maximum performance at minimum costs. To make an educated decision on when to scale in or out, there are rules defined in a Scaling Policy that determine what action to take when the target system is experiencing fluctuation in workload. To examine the efficiency of different Scaling Policies an experimental environment was built. This environment can simulate scenarios where the workload experiences an unpredictable burst.
The results show that a Scaling Policy which uses an Exponential Moving Average scores slightly better than with a Simple Moving Average given the workload fluctuations that satisfy the Load Profiles. Furthermore it shows that a target range of 60/85 has at least a 10% improvement on the efficiency over a target range of 50/75.|
|Document type|| scriptie master|
Use this url to link to this page: http://dare.uva.nl/en/scriptie/394699
Contact us about this recordNotify a colleague
Add to bookbag