Improving Resource Allocation Methods Based on Service Level Agreements in Cloud Computing Environments

Master thesis

Information technology engineering

Information systems management trend

Abstract

Cloud computing is a new computing model in which software, hardware, infrastructure, platform, data and other resources are provided to cloud users virtually and as a service, on demand and through the Internet by cloud providers. This model is based on pay-per-use, that is, users pay only for the resources they use. Among the topics raised in these environments, we can mention things such as security, resource management and scalability. Cloud environments provide countless resources to users, and users can reduce or increase the amount of resources they need, so resource management is one of the most important issues in cloud computing. Providing service in the cloud is based on a service level agreement (SLA). A service level agreement is a tool for meeting non-operational needs such as quality of service between cloud providers and users. In order to achieve SLA, providers must be able to schedule resources and deploy applications in ways that meet SLA goals and thereby achieve customer satisfaction. In case of violation of SLA, the provider must pay a fine to the user. Therefore, what is very important for providers is to resolve customer service quality issues in order to get more customers and reduce fines and increase profits. To meet SLA requirements, providers may want to set up a separate virtual machine (VM) for each request (static allocation), in this case, although the service level needs may be met, hardware resources are wasted, which ultimately leads to increased costs for the provider. To solve this problem, multi-tenancy methods were proposed, in which a VM can service several requests, but care must be taken to ensure that the SLA of different requests is met, and if methods can be provided that meet this importance, higher efficiency can be achieved (dynamic allocations). Therefore, in this research, a multi-tenant scheduling method based on SLA is proposed with the aim of efficiently deploying requests on VM. In this method, an SLA parameter called time limit and a parameter as priority for user requests and an SLA parameter called cost for the provider have been defined, and the results have been evaluated based on cost. The cost is the amount of fine that the provider pays for violating the SLA. Cloudsim tool has been used for implementation and to evaluate the proposed method (dynamic allocation: including migration, priority and SLA) it has been compared with two simple allocation methods (without considering migration, priority and SLA) and static allocation (without considering migration and respecting priority and SLA). The results show that in the proposed method, the cost is lower than the other two methods, because due to the consideration of priority and SLA, non-exclusivity of the VM by the request and the possibility of migration of requests, more requests reach the resources they need and can finish according to their deadline, and this means the effective deployment of requests in VMs and optimal use of resources.

Keywords: cloud computing, virtual machine, resource allocation, service level agreement

Chapter One

Introduction

1-1 Problem Definition

Before the advent of computers, to find the required information, one had to physically refer to certain references, and this caused a lot of time and cost to be lost. With the spread of computers, the process of finding and using information became easier, and with the advent of the Internet and then web-based services, a valuable development was created in the field of information technology. With the development of information technology, there was a need to perform computing tasks anywhere and anytime, and there was also a need for organizations and individuals to be able to perform their heavy computing tasks - without having expensive hardware and software - through service services. Cloud computing[1] has been the latest response of information technology to these needs. Cloud computing is a contract that allows access to applications, hardware, software, data, platforms and infrastructure through the Internet and based on a payment model.Cloud computing is a contract that provides access to utilities, hardware, software, data, platform and infrastructure through the Internet and based on the payment model based on the amount of usage2, that is, in these environments, many services are virtualized and users request only the resources they need and only pay for them. Therefore, in these environments, there are three main factors that include the provider, user, and resources. The goal of this emerging computing paradigm is things such as scalability, availability, throughput, resource usefulness, resource management, and security, and it faces challenges to achieve these goals. With the increasing popularity of cloud computing, research centers and enterprises began to outsource their computing needs to cloud services. Clouds are typically very large-scale virtualized data centers 3 that host large numbers of servers. Although this virtualized infrastructure has many advantages, including scalability according to resource demand, there are still issues that prevent its widespread adoption. In practice, in order for these environments to achieve commercial success, it is necessary to provide better and more accurate service quality guarantees. These guarantees, which are documented as service level agreements, are very vital, because only then will users be confident in outsourcing their work to the clouds. Therefore, compliance with SLA5 objectives in order to obtain customer satisfaction and increase the provider's profit are considered important issues in these environments. Provisioning plays a key role in ensuring that cloud providers fulfill their commitments to users.  Providing resources effectively is a challenging issue in cloud computing environments due to the dynamic nature of requests and the need to support heterogeneous applications with different performance needs, in other words, the differences in nature between different workloads make it difficult to provide resources. There are various methods for sourcing, including methods based on service level agreements, market-oriented methods[2], law-based methods, and methods based on reducing energy consumption. Each of the aforementioned methods focuses on a specific issue, for example, the goal of methods based on reducing energy consumption - as the name suggests - is to provide algorithms that can reduce the amount of power consumption in the cloud [9-11] or the focus of market-oriented methods is to increase the provider's profit [3] and [18].               

               It can be considered that the reduction of energy consumption and the increase of profit are considered as a kind of SLA. That is, these methods can also be considered as SLA-based methods, in which the SLA parameter is considered to reduce energy consumption in one and increase profit in the other. In many methods, parameters such as execution time [15-17], response time and completion time [25] and [28] are considered as SLA. What is important is that if the customer's desired goals are not met, he does not entrust the performance of his work to the cloud, and the provider must provide guarantees to obtain customer satisfaction. Therefore, in this research, resource allocation based on SLA has been investigated. If the SLA required by the users is met, the users will accept these environments more and entrust their work to the clouds with more confidence, and as a result, many expenses for the purchase of hardware, software and other equipment are removed from the organization's list of expenses, and instead of focusing on the purchase and adjustment of the required hardware, software, platform and infrastructure, the organizations focus on their strategic goals. A large number of users, providers need to provide different services in ways that meet the quality expectations of users, therefore it is very difficult to provide resources in the cloud, and the resource provider must calculate the best hardware/software configuration in order to guarantee QoS goals in order to maximize the usefulness and effectiveness of the environment. of QoS between the provider and the user) are provided. Achieving QoS is very important for SLA compliance, because the level of user satisfaction depends on the SLA compliance.