Efficient, Scalable and Survivable Design for next generation Virtualized Data Center for Cloud Services

点击翻译按钮获取中文摘要

Cloud Computing is becoming the mainstream paradigm, as organizations, both large and small, begin to harness benefits. Cloud computing gained its success for giving IT exactly what it needed: The ability to grow and shrink computing resources, on the go, in a cost-effective manner, without the anguish of infrastructure design and setup. The ability to adapt computing demands to market fluctuations is just one of the many benefits that cloud computing has to offer, this is why this new paradigm is rising rapidly. According to a Gartner report, cloud services are expected to grow to 131 billion dollars worldwide in 2013. With this massive growth, the performance of the underlying infrastructure is crucial to its success and sustainability. Currently, cloud computing heavily depends on data centers for its daily business needs. In fact, it is through the virtualization of data centers that the concept of “computing as a utility” emerged. However, data center virtualization is still in its infancy. Recently, NSERC has created a Strategic Network for Smart Applications on Virtual Infrastructure (SAVI) whose objective is to address the design of future applications platform built on a flexible and evolvable infrastructure. Among SAVI’s objectives is helping advance Canada's Digital Economy Strategy by strengthening the IT (Information Technology) industrial base through active participation in the research program. This initiative underscores the importance of research on cloud networks and network virtualization and the impacts they will have on Canadian industries. With this initiative, however, a plethora of open research issues and challenges related to data center virtualization can be identified, including but not limited to, optimized topologies and protocols, embedding design methods and online algorithms, resource provisioning and allocation, data center energy efficiency, fault tolerance issues and fault tolerant design for cascaded correlated failures, improving service availability under catastrophic disaster conditions, etc. Therefore, this proposal will attempt to elaborate and address key research challenges and problems related to the design and operation of efficient virtualized data centers and data center infrastructure for cloud services. In particular, I plan to investigate the problem of scalable traffic management and traffic engineering methods in data center networks and present both mathematical formulation for the (offline) problem as well as online methods/algorithms to enable quality of service (QoS) provisioning. I also plan to study the problem of survivable virtual network embedding through novel coordinated design methodologies to achieve higher revenues for network operators while providing guaranteed service continuity for cloud services. Given the complexity of this problem, we plan to use large scale optimization methods for addressing the offline design problem and novel algorithms for obtaining online solutions. Now, one of the major issues for network operators is to maximize their profits; this implies squeezing the most benefits of their network investments and capacity to be able to accommodate more services. Network coding is a framework that allows network operators to maximize the traffic flows in their network without requiring any upgrade to the infrastructure. To get the most benefits, network coding aware embedding solutions must be devised. This is however another challenging problem that has not been addressed in prior work. This proposal plans to investigate this problem through novel embedding solutions as well as offline and online methods to characterize the gains.