MultiCloud Tournament: A cloud federation approach to prevent Free-Riders by encouraging resource sharing

Abstract

Organized in associations, Cloud Computing Providers became able to negotiate the acquisition of resources from other providers at more attractive prices, while allowing the marketing of their own idle resources to the other members of the respective Inter-Cloud. However, Inter-Cloud associations are characterized as resource-sharing environments, which allow for the presence of malicious providers called Free-Riders. Free-Riders have predatory behavior, seeking to serve only their own interests and consequently harming other providers within the environment. Mechanisms to avoid Free-Riders, as well as to keep resources supply × demand balanced, are needed in Inter-cloud environments such as Cloud Federations. For example, the absence of a resource consumption control mechanism can cause a Free-Rider to consume resources in a way that is disproportionate to that offered to Inter-Cloud, referring to the problem described previously. In view of the above, this paper presents a new Inter-Cloud called MultiClouds Tournament (MCT) focused on addressing the standalone cloud providers limitations. This Inter-Cloud is inspired by the format of a soccer tournament and based on the main features of Cloud Federations. We assessed the MCT using a real workload, showing it offers an architecture that adds to the characteristics of the federations, being effective in the management of CSPs and robust in relation to Free-Riders prevention and the main strategies CSPs.

Introduction

The growth and effective use of the Internet for a wide range of purposes (commerce, leisure, etc.) has leveraged research into new technologies for offer service-oriented computing: Service Oriented Architecture (SOA) (S and Reference architec, 2012) and Grid Computing (Berman et al., 2003; Foster et al., 2008) are examples of them. Grid computing has made it possible to obtain computational resources in a way similar to the acquisition of services. However, the lack of control on the part of the users over the domains where their workloads were being processed, the scientific appeal of the solution, among other characteristics, have discouraged the use of Grid Computing in a commercial scope. This paved the way for similar, more widely-focused technologies such as Cloud Computing.

With a commercial appeal and leveraged by virtualization technologies, Cloud Computing focused on offering computing as services to interested parties under an innovative business model for the infrastructure offer segment. With maturation of the Cloud Computing paradigm, some limitations have been exposed by Cloud Service Providers (CSPs), especially the difficulty of maintaining the supply elasticity in the delivery of resources (Buyya et al., 2010; Toosi et al., 2011) and problems in sustaining the quality of services offered due to regional usage behavior (Buyya et al., 2010). Such limitations have promoted the extension of the paradigm allowing the grouping of multiple clouds associations, mainly in non-profit scenarios (e.g. EGI (European Grid Infrastructure (EGI1), 2019)). These associations were named Inter-Clouds (Use cases and funct, 2010), where Hybrid Clouds (Bittencourt and Madeira, 2011), Multi-Clouds (Kurze et al., 2011), Sky Computing (Keahey et al., 2009) and Cloud Federations are the main representatives. Inter-Clouds address a set of properties of the Cloud Computing paradigm that monolithic CSPs are not adept at offering. In addition, the Inter-Clouds opened up a new resource market, which made it possible to increase the revenues of providers (Falcão et al., 2015; Villegas et al., 2012).

Organized in associations, providers became able to negotiate the acquisition of resources from other providers at more attractive prices, while enabling the marketing of their own idle resources to the other members of the respective Inter-Cloud. However, Inter-Cloud associations are resource-sharing environments, which allow for the presence of malicious providers called Free-Riders. Free-Riders prioritize their own revenue and are not aware of the balance of the $supply\times demand$of resources available in the environment they are in (Hammoud et al., 2019). These Free-Riders hinder associations and may discourage the use of these multi-clouds infrastructures from other providers with good intentions, which are aware of the availability of resources in the environment. A direct consequence of the presence of Free-Riders is the reduction of resources in the environment, so there may be inflation in the prices of acquisition of the remaining resources for example. Even associations that have certain control mechanisms, such as Cloud Federations regulated by an internal contract, are not able to totally preventing the presence of these Free-Riders.

Mechanisms to avoid Free-Riders, as well as to keep resources $supply\times demand$balanced resources, are needed in environments such as Cloud Federations. Such mechanisms should be able to perform life-cycle management of CSPs by providing incentives to those providers that perform well within the environment - in this context it can be described as the supply and consumption of resources in the same proportion throughout the time. For example, the absence of a resource consumption control mechanism can cause a Free-Rider to consume resources in a way that is disproportionate to that offered to Inter-Cloud, referring to the problem described above. This behavior decreases the amount of resources in the environment, making the environment unattractive for well-intentioned CSPs. In addition, these management mechanisms must maintain the environment in line with the contract that defines the federation itself.

In view of the above, this article proposes a new Inter-Cloud called MultiClouds Tournament (MCT) focused on addressing the limitations exposed. This Inter-Cloud is inspired by the format of a soccer tournament and based on the main features of the Cloud Federations that will be presented throughout the text. This work presents the latest architecture, the formalization of all defined components, and validation experiments, which were preliminarily introduced in (Assis and Bittencourt, 2017). In this work, the description of the MCT elements was deepened: players, statute, referee, and divisions. In addition, they were presented with new components of the MCT that were being built during the development of the architecture: pre-season, tournament operation, and traffic rate. Moreover, the formalization of both the problem of interest and the Free-Riders behavior are introduced. Finally, experiments focused on Free-Riders’ strategies were presented.

The main contributions of this article are:

• Describe the Cloud Federation properties and the problem inherent to it (Tragedy of Commons);

• Define Free-riders and analyze their main strategies;

• Propose a MCT Inter-Cloud, which is able to avoid Free-Rides and improve the cloud providers association.

The remainder of this document is organized as follows: Inter-Clouds and Cloud Federations are presented in Section 2. In Section 3 is the main motivation of the proposal. Then, in Section 4 the MCT is described. The approaches used by MCT to prevent the mainly Free-Riders strategies, the experimental validation of the architecture, and the conclusion are described in Sections 5 Preventing Free-Riders strategies, 6 Validation and results, 7 Conclusion respectively.