Cost-aware Optimized Virtual Machine Placement through Cloud Broker in Multi-cloud Architecture Md. Golam Rabiul Alam, Choong Seon Hong Department of Computer Engineering, Kyung Hee University robi@networking.khu.ac.kr, cshong@khu.ac.k Abstract Cloud services through cloud brokers are upcoming services acquisition approach for cloud service consumers, where cloud broker acts as an intermediary between cloud service providers and service consumers. From the viewpoint of service charge, the IaaS cloud service providers impose higher services charges for their services. As the service charge varies dynamically in respect to cloud service providers, in respect to requested cloud resources and timestamp. In this paper, we propose a cost- aware selection of cloud providers and their virtual machines (VMs) as well from the broker perspective to provide cost effective services to consumers. We consider multi-cloud environment for cloud service provisioning and apply portfolio optimization technique to purchase services from cloud service providers. 1. Introduction The charges of various IaaS provider [1], which provides a multi-tenant on-demand multi-cloud computing environment [2], are varies upon user demands, contracts, contract periods, location, service availability and requested virtual machine instances etc. As for example, windows azure charges $0.12 for per hour CPU usages and amazon EC2 charges minimum $0.14 for per hour CPU usages. The objective of this paper is proposing a method of cost effective service selection i.e. cost effective VM placement from brokers’ perspective. 2. Related Works The optimized placement of virtual machines in cloud brokering architecture is proposed in reference [3]. The paper presented very detail architecture of cloud service broker. The optimized selection of virtual resources of cloud brokers through cloud scheduler was one of the primary objectives of the paper. The authors’ of the paper didn’t consider dynamic cloud scheduling and dynamic pricings to place virtual machines dynamically. The cost-effective deployment of computing clusters in multi-cloud infrastructure is presented in reference [8]. They provide analysis on the viewpoint of performance and cost. The proposal is only for loosely coupled many task computing (MTC) applications. The proposal may not perform well for tightly coupled MTC applications, where facts are highly interdependent and synchronization among the computational units are necessary. To maximize the revenue of cloud data center, a joint or coordinated VM resource provisioning and maintenance scheduling method is proposed by the authors of reference [4]. They formulate the problem as an Integer Linear Programming problem and then transformed it to equivalent problem to obtain linear programming relaxation solution, then they apply LIST rounding algorithm towards final approximate solution. 3. System Model The considered system model of cloud based service provisioning is presented in figure 1.The system model consists three stuffs, these are service consumers, cloud service broker and cloud service providers. The service consumers are the end users. The cloud service providers are the infrastructure and compute service vendors, who sales IaaS services with dynamic charges. The cloud service brokers are the intermediary entity, who acts as an agent of cloud service providers. The responsibilities of cloud service broker are to accept service request from service consumers, aggregate services of cloud service providers, create virtualized cloud resources, and ensures optimized cloud services through its optimizer and virtual resource manager. The proposed 2014년 한국컴퓨터종합학술대회 논문집