Dynamic Re-provisioning of Virtual Data Centers (VDCs) in Cloud Data Centers

Project: Research

View graph of relations

Description

With the rapid development in recent years, cloud computing technologies have enabled more efficient, flexible and collaborative computing models. In cloud computing, one of the most important service models is infrastructure-as-a-service (IaaS), which allows a large number of application service providers to share the same physical infrastructure.To implement IaaS, a cloud data center usually hosts hundreds of virtual data centers (VDCs) with various sizes and service level agreements (SLAs). With the dynamic creation, resizing, and termination of VDCs, there are many open issues and major challenges to efficiently and dynamically provision and re-provision VDCs in a cloud data center. In this proposal, we will address these timely and challenging issues. In particular, we will study the following scenarios.Firstly, the dynamic creation, resizing, and termination of VDCs often result in the fragmentation of cloud resources. To optimize the system performance, it is essential to defragment resources by VDC re-provisioning. Nevertheless, it is very challenging to understand and manage the tradeoff between the re-provisioning cost and the performance improvement of both the cloud data center and VDCs.Secondly, to ensure high availability of the cloud infrastructure, a cloud data center often needs to conduct periodical software and hardware maintenance. During the period of maintenance, some affected VDCs must be re-provisioned to ensure the smooth running of such VDCs. Thus, one big challenge is to find the optimal timing to conduct the required maintenance activities before their deadlines and to minimize the disruption to VDCs.Thirdly, following a power outage, a cloud data center will be running under a limited power budget before the power supply comes back in full. In this scenario, a major challenge is how to minimize the violation of SLAs. This involves integrated backup power scheduling, selectively downgrading of VDC performance, and VDC re-provisioning.In summary, this project will investigate how to optimize dynamic VDC re-provisioning, whenever it is appropriate or mandatory, to achieve high utility of the cloud data center and to provide the best possible SLAs. We expect to develop solutions to empower the cloud service providers to make the appropriate decisions under various situations, i.e., to balance the tradeoffs between the cost of dynamic VDC re-provisioning and the system performance. To the best of our knowledge, this is the first attempt to systematically study VDC re-provisioning in a cloud data center.

Detail(s)

Project number9041787
Grant typeGRF
StatusFinished
Effective start/end date1/07/1230/05/16