Authors:
Shoubin Kong
;
Yuanping Li
and
Ling Feng
Affiliation:
Tsinghua University, China
Keyword(s):
Cloud computing, Infrastructure-As-A-Service (IaaS), Virtual machine (VM), Resource configuration, Performance, Expenses, Multiple objective optimization.
Related
Ontology
Subjects/Areas/Topics:
Cloud Applications Performance and Monitoring
;
Cloud Computing
;
Information and Service Economy
;
Platforms and Applications
;
Service Strategy
;
Services Science
Abstract:
IaaS is a popular cloud computing service paradigm based on virtualization technology. In an IaaS cloud environment, the service provider configures VMs with physical computing resources (e.g., CPU and memory) and leases them to IaaS customers to run their applications. The customers pay for the resources they use. Such a pay-as-you-go charging mode brings about a few critical concerns about the expenses paid and the performance received. From the standpoint of cloud customers, such concerns as minimizing the expenses while ensuring the performance, optimizing the performance within the budget limit, compromising the expenses and performance, or balancing performance of applications running on different VMs, etc. thus arise. For the IaaS provider, how to reasonably configure VMs so as to meet various requirements from different customers becomes a challenge, whose solution influences the acceptance of IaaS in the future. In this paper, we address this problem and present a weighted m
ultiple objective optimization approach for flexible control of expenses and performance in an IaaS cloud environment. We focus on database applications, consisting of various queries to be executed on different VMs. A genetic algorithm is implemented based on a fine-grained charging model, as well as a normalized performance model. Experiments have been conducted to evaluate the effectiveness and efficiency of our approach, using TPC-H queries and PostgreSQL database in a simulated cloud environment.
(More)