| Large-scale Internet applications provide service to end users by routing service requests to geographically distributed servers, which may be located at different data centers. Two concerns exist in service provisioning by the data centers. One is that the users require to experience low latency while accessing data from the data centers. The other is to reduce the power consumed by network transport and the servers in the data centers. For the former, develop the data which the users need in the data center which is close to the user is one of the effective ways to reduce the latency from users to access data. For the latter, data centers and networks consume huge energy cost, develop appropriate number of data to reduce the energy consumption of data centers and networks. How to develop appropriate number of data to reduce the latency, energy consumption of data centers and networks is the main issue of this thesis.The main contribution of this thesis are as follows. This thesis builds an energy consumption model of the data from users to data centers through networks and formulates a green latency-aware model of single copy data deployment problem. This thesis proposes GLDD algorithm to search a server in a data center to accommodate each data item. The simulation results have shown that the proposed algorithms are effective in terms of the reduction of data center power consumption, as well as the tradeoff among the latency of data access, the power consumption of the network transport and the data centers.This thesis also forms a green latency-aware model of multi-copy data deployment problem and proposes algorithm Merge to combine data items into large data groups, and then proposes two heuristic algorithms to find a server in a data center to hold each data group. The simulation results have shown that the proposed algorithms can reduce the total cost of the latency of data access, the power consumption of the network transport and the data centers. |
PDF Full Text Request