Multi-layer quality-aware (MULQA) cloud framework

Abstract

In the past few years, the popularity of cloud-based solutions in the IT domain has been increased significantly as the consequence of the industry shift towards IoT, super-fast computer networks and notably the benefits of emerged cloud computing. However, this leads to many technical challenges such as optimizing the infrastructure for heterogeneous applications especially the quality sensitive types, and issues toward addressing different quality attributes simultaneously. In this research, we propose MULQA, an autonomic framework that monitors and estimates the quality metrics in physical, infrastructure, platform and software layers of an open source cloud system, and ensures the quality of the targeted metrics by triggering appropriate actions. MULQA is a novel approach providing such framework which targets different quality metrics in all layers of the cloud. During this thesis, we describe MULQA framework where the analyze module, predicts the violation status of the quality metrics and this predicted information will be used to create events for the finite state machine of the planning platform. This control mechanism consists of Normal, Warning and Transition states. Warning state is used to prepare the cloud for the transition state, while transition state prevents the violations and brings back the system to the normal state. Being a modular framework, MULQA provides generic functionalities and modules that can be selectively changed by additional user-written code, which can be used to test proposed algorithms for Monitor, Analyze, Plan and Execute modules. MULQA framework is built to overcome the challenges in providing a loosely coupled system which can be easily distributed and customized through an API. Furthermore, this framework is compatible with Openstack architecture and is able to monitor and control the components that the cloud middleware doesn’t have access to. The use-case in this thesis, is a three-tier Web application which is deployed with Openstack. Experimental results of the tests which focus on the performance QA, show that MULQA can increase the success rate of requests sent by 32%, 69% and 94% for request concurrency numbers of 200, 500 and 1000 in order. Moreover, throughput has been improved five times with low impact on the CPU utilization.

Date	9 Jun 2017
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Mohamed Cheriet (Supervisor)