Research On Energy Efficient Virtual Data Center Embedding Approach

As the core infrastructure of IT systems,data centers are clusters of servers connected by a large number of switching devices,providing a large amount of computing,storage and network resources.In the process of resource scheduling,traditional data centers face many problems in terms of efficiency,reliability and security,and the virtual data center can effectively solve the above key problems.As a typical feature of new data centers,it allows multiple users to share different types of resources in the same physical data center without being affected by each other.In virtual data centers,the most important issue is how to allocate corresponding node and link resources for virtual data center requests,that is,virtual data center embedding problems.At present,this problem has attracted lots of attentions from domestic and foreign scholars.However,through analysis of the problem,we find that prior studies focus on how to improve the revenue of physical data center providers,while ignoring the energy issue.Regarding this problem,we have conducted in-depth research and proposed innovations as follows:(1)Proposing an energy-aware single-domain virtual data center embedding algorithm,which can reduce the energy consumption of physical data centers.In the virtual data center embedding scenario,the existing research only regards the oprerating revenue of physical data centers as the main evaluation metric of the virtual data center embedding,while ignoring the energy consumption problem of physical data centers in the virtual data center embedding.The problem has seriously affected the profit of physical data center providers.In order to optimize the energy consumption of the physical data centers,considering the context of the virtual data center embedding problem,this thesis first studies the energy consumption model for virtual data center embedding.Considering reducing the energy consumption of physical data centers and keeping virtual data center embedding request acceptance ratio,this thesis proposes two energy-aware virtual data center embedding algorithms,which are based on label propagation technology and particle swarm optimization,respectively.Experiment results show that the algorithms above can significantly reduce the energy consumption of physical data centers while maintaining the operating revenue of physical data centers.(2)Proposing a cost-aware cross-domain virtual data center embedding algorithm which can reduce the energy consumption of the physical data centers.Prior studies on virtual data center embedding mainly focus on single-domain while ignoring the practical scenario that physical data centers are cross-domain deployed.For the cross-domain virtual data center embedding problem,this thesis makes a comprehensive consideration about the dual factors of electricity price and energy consumption while ensuring the acceptance ratio of embedding requests,then leverages the difference of electricity price in different regions and different times,and builds the electricity cost model for cross-domain scenario based on the single-domain virtual data center embedding problem.Then this thesis proposes a cost-aware cross-domain virtual data center embedding algorithm to further optimize the electricity cost of the data center provider.The experiment results show that this proposed algorithm can save a lot of electricity cost for physical data centers while keeping the acceptance ratio for virtual data center embedding requests.(3)Proposing an energy-aware virtual data center migration algorithm to further reduce the energy consumption for physical data centers.Existing research focuses on how to perform a single virtual data center embedding to improve virtual data center request acceptance ratio.There is no dynamic adjustment of the embedding scheme.However,with the arrival and departure of virtual data center embedding requests,the resource of the physical data centers will be allocated and released,the distribution of resource will change dynamically along with time,and the optimal solution obtained by the previous embedding algorithm may become a sub-optimal one.This thesis leverages the migration technique to migrate the virtual data centers,and proposes an energy-aware virtual data center migration algorithm based on Ant Colony Optimization.The algorithm consolidates the virtual data centers and further reduces the energy consumption in the virtual data center embedding process while ensuring the acceptance ratio of virtual data center embedding requests.The experiment results show that this proposed algorithm can reduce energy consumption significantly while keeping the acceptance ratio for virtual data center embedding requests.