A Survey on Efficient Coalition Formation Game in Cloud Federation of Applications

Recently, the growing demand for various multimedia services and applications has imposed great challenges in terms of real-time processing, storing, analyzing and sharing of large media data. Federation among media cloud providers can be a promising solution for effectively hosting multimedia applications and provisioning of Quality of Service (QoS)assured services. In this paper, we address the problem of efficient federation formation mechanism among media cloud providers (CPs) that motivates them to cooperate to fulfill the dynamic resource demands of users for running multimedia workloads. We design a novel media cloud federation formation mechanism based on cooperative game theory. By collaborating to trustworthy providers, our proposed method looks for building federations to minimize the penalties due to possible violation of service quality by untrustworthy providers, and thereby, generate more profit. Simulation results demonstrated that the media cloud federation obtained by the proposed mechanism is stable and provide higher profit for the participating CPs. I.Introduction Recently, multimedia is emerging as a popular service that involves in live broadcast, on demand videos, games, video conference and others. Using wireless or wired networks, various users can ubiquitously access and share diverse media services. As the media service demand is increasing day by day, the media traffic volume for all types of services will also increase in near future. This speedy growth of the demand for multimedia services has led to great challenges in terms of large volume of data processing, storing, analyzing and sharing among millions of users with high interactivity [1]. In order to address the above issues, we propose to utilize the federation paradigm [2][3][4] among various media cloud providers to serve dynamic and fast growing multimedia services. Cloud federation presents a promising approach for effectively hosting multimedia applications and provision of QoS-assured services. Leveraging infrastructure provisioned by multiple partnership media cloud service providers, we solve problems like dynamic resource allocation and improvement of user experience. The federation IJCSIET--International Journal of Computer Science information and Engg., Technologies ISSN 2277-4408 || 01062016-007 IJCSIET-ISSUE6-VOLUME2-SERIES2 Page 2 creates a pool of virtualized resources which are offered to users as different types of VM instances. The CPs in a federation cooperate in order to provide the resources requested by users with guaranteed QoS. As a result forming cloud federations helps to achieve greater scalability and performance. It may also provide the resources at lower costs[5]. However, one of the major challenging issues in this environment is how to define an effective federation formation mechanism among CPs that motivates them to cooperate to fulfill the resource demands of users [5][6][7] for big data workloads. This issue is very important since it determines how much revenue each CP gets when participating in a federation by proving a certain number of VM resources. In addition, the fairness is also an important issue to be considered which ensures that each CP should gain a revenue based on the amount of VM resources contributed to the federation (Gain-as-you-contribute in short). The stability of a federation is also important and needs to be analyzed for finding whether it is profitable for a CP to work with other CPs in the federation or not. Currently, several strategies [5][6] are present in the literature for forming cloud federations. Among these strategies, gametheory-based federation formation mechanisms are becoming popular[5][6][8]. However, all of these mechanisms mostly focus on forming federation based on the maximum individual profit gained by the CPs in a federation. The profit of any CP varies based on the revenue it can gain by running various types and number of VM resources and the cost of providing those resources. None of these approaches take into account the risk of selecting unreliable collaborating CPs in a federation that may result in additional penalty cost due to SLA violation and finally affect the other CPs reputation. Therefore, we argue that a trust model is also necessary to find the most promising collaborators for forming cloud federation. In this paper, we model the federation formation among CPs as a coalition game, where CPs decide to form a federation to allocate various types of VM instances dynamically, based on usersa ̆A ́Z requests with QoS guarantee. The coalition game is modeled as a hedonic game [5][6] whereby each CP can order and compare all the possible coalitions it belongs to based on its preference model satisfying fairness and stability properties. We derive a trust model specifically designed to assist a CP in making its own preference model about federation formation. We analyze the properties of our proposed cloud federation formation game and perform extensive experiments to investigate its properties. The paper is organized as follows. We present the overall system model and mathematical IJCSIET--International Journal of Computer Science information and Engg., Technologies ISSN 2277-4408 || 01062016-007 IJCSIET-ISSUE6-VOLUME2-SERIES2 Page 3 problem formulation. We describe the proposed cloud federation formation game in detail. In section 4, we evaluate the effectiveness of the proposed game in a cloud federation environment, and finally, in section 5, we present our conclusions and future work. II. Mathematical Formulation We first describe the system model consisting of a set of media CPs, several trust evaluation agent (TEA), some cloud users and a broker as a mediator. Each media CP has a TEA which provides trustrelated information services to other CP’s TEAs and users [12]. One TEA can represent multiple CPs, and one CP can delegate multiple TEAs. A TEA can derive the objective trust of CPs for each QoS factors from the trust information received from monitoring information process agents and users. It also collects the trust feedback ratings sent by other TEAs of other CPs for services that they have used, in order to build up the subjective trust of each service. Furthermore, TEAs can also interact with each other in order to exchange and propagate trust information. The broker in a federation is a trusted third party responsible for receiving cloud users VM requests, motivating various CPs to cooperatively supply required amount of VM resources in the federation, receiving the payment from users, and dividing the profit among CPs in the federation. III. Cooperative Cloud Federation Formation Game In this section, we introduce our proposed media cloud federation formation game.In this game, a broker announces a price or revenue rate along with the total amount of VM resources required by the media cloud users. Accordingly, a set of available CPs forms a coalition to cooperatively fulfill the requested user demands. However, there are many different coalitions can be formed, each one differing from the other ones in terms of the profit they bring to their members. Therefore, the proposed game model will help to investigate the stability of different federation structures and find the best federation model that maximizes each CPs profit by considering trustworthy collaborator CPs. We consider a coalition formation cooperative game with transferable utility [5][6]. The proposed game also satisfies the two main properties, fairness and stability, which are very important for any CP to decide which coalition to join. The profit obtained by a federation should be fairly divided among the participating CPs. In addition, a federation should be stable, that is, the participating CPs should not have incentives to leave the federation. In the following IJCSIET--International Journal of Computer Science information and Engg., Technologies ISSN 2277-4408 || 01062016-007 IJCSIET-ISSUE6-VOLUME2-SERIES2 Page 4 sections, we describe the proposed game in detail. IV. Federation Formation Process In the proposed media cloud federation formation game process, a preference relation needs to be defined so that each CP can order and compare all the possible coalitions it belongs to, and hence it can build preferences over them. Unlike existing works, we consider that the preference of a CP to join/leave any federation will depend on the maximum profit it can gain from that federation with minimum risk of penalty costs by selecting trustworthy collaborators. The federatoin formation algorithm has two important propertiesconvergence guarantee and Nash-stability, which are proved in [green cloud]. In the next section, we show how to calcualte the trust value of any CP when forming a federation. V. Performance Analysis In this section, we present our evaluation methodology and simulation results to show the effectiveness of the proposed cloud federation formation game. In our game model, we consider that each CP will calculate its total cost of providing VM resources based on the production cost of VMs, running energy cost of servers and risk or penalty cost due to various trust levels to other CPs. We assume that the trust level of different CPs exists based on our trust model executed by various TEAs of CPs. Our simulation environment consists of the participation of 8 independent CPs. Each CP consists of a number of hosts (servers), VMs and a trust value picked randomly from the range presented. To compare our proposed method with other mechanisms, we have adopted two other state of the art approaches. One [6] of which considers cost in terms of production and running energy cost of servers but does not take into account the penalty costs due to trust; while the other one [5] considers neither the penalty nor the energy cost when deriving its profit from the federation. For the sake of simplicity in referring to them, we refer to the first approach here as Federation game with energy cost (as because it considers the production costs and the running costs) method and the second approach as Federation game without energy and trust penalty cost. the individual profits made by the different CPs. Here the federations formed by the proposed approach are found to make higher total profits than the other two approaches. This is because the other two approaches do not consider the costs related to running the servers and trust of the CPs which minimizes the expected profit. Here we can also notice that some CPs have made profits in the PR and P approaches while the proposed approach have not picked IJCSIET--International Journal of Computer Science information and Engg., Technologies ISSN 2277-4408 || 01062016-007 IJCSIET-ISSUE6-VOLUME2-SERIES2 Page 5 them in the federation. Actually, the proposed method has avoided them in the federation because such CPs may not lead to the expected profit due to their trust values. Considering the penalty costs which may incur due to the low trust value, the proposed method may not always pick a CP in a federation even though it promises to make high profit.