abstract.tex

%!TEX root = thesis.tex
\begin{spacing}{1}
When writing code for a cloud computing software platform testing is a must. Unit testing and integration testing can help the developer know if his/her code will cause bugs in the overall system. However, cloud platforms are very complex pieces of software, with many many ``moving pieces''. There are times when we will want to write code to change the system's behavior ---for example, to include new Virtual Machine placement or consolidation techniques. In these cases it is not enough to hunt for bugs. We need to know how the new code will behave within the system as a whole. To do this we need to be able to run complex simulations, and to collect and analyze the obtained results. However, it is difficult for a developer --who is often provided with limited amount of hardware resources-- to experiment code in an entire cloud system which needs tens of physical machines to be hosted. Virtualizing the system is thus compulsory.

Our solution is aDock, a modular system that leverages Docker's lightweight virtualization techniques and OpenStack --our reference open-source cloud computing software platform-- to allow users to deploy extremely lightweight cloud systems, to run simulations, to store the system's output, and to display it in real-time thanks to a friendly user interface.

As further contribution we developed a Virtual Machine Consolidation service for OpenStack, and tested it with four different consolidation algorithms. Virtual Machine Consolidation is one of the topics of primary importance in nowadays cloud systems. Consolidation is supposed to be an intelligent and efficient strategy for resource allocation, in order to make the most of available hardware and, thus, save energy. Energy saving, in fact, has become an urgent and important problem in data centers due to their growing energy greediness.

In order to evaluate our solution we used aDock to deploy an OpenStack system, and to benchmark and compare the proposed consolidation algorithms. Thanks to aDock we were able to deploy the system in a reasonable time ---even on our laptops. The simulation results demonstrated that the four consolidation algorithms brought various degrees of improvements to the system's resource allocation.
\end{spacing}