Research On Power Saving For Inter-Datacenter Optical Interconnection Network

With the rapid development of cloud computing,datacenter technology and facilities that provide information storage and processing,have emerged as the key infrastructure for supporting essential Internet functionality and services.At the same time,the requirements in terms of inter-datacenter communication to support data backup and data synchronization between different datacenters trend to be higher,which call for appropriate network technologies to interconnect datacenters effectively and efficiently.The elastic optical network can dynamically schedule the exact number of frequency slots according to the actual size of the traffic to establish lightpath,so as to serve multi-granularity traffic to achieve transmission effectively.Therefore,the elastic optical network is regarded as the most appropriate technology to interconnect the inter-datacenter.However,with the explosive growth of datacenter traffic,the energy consumption of datacenter interconnected by elastic optical network becomes more prominent.For the specific datacenter traffic,energy efficiency design has received significant attention within the research community recently.The main contents and innovations of this paper are as follows:To effectively reduce the energy consumption for datacenter interconnected by elastic optical network,an energy saving strategy based on software defined networking is proposed.The strategy is designed to reduce the huge switching power of lightpath termination and re-establishment.That is,the software defined networking technology and back propagation neural network optimized by particle swarm model are used to predict the traffic amount of each lightpath in the network firstly.Then,for the lightpath on which the traffic transmission has been completed and is about to be removed,a holding time is set for the lightpath according to the prediction result.By setting the holding time for the idle lightpath,it can avoid the re-establishment of the lightpath between the same source destination node pairs,eliminating the switching power consumption.Compared to the strategies which do not consider the switching power,the simulation results show that the proposed strategy can effectively reduce the total energy consumption.And it will not lead to the bandwidth blocking ratio increase greatly.The above energy saving strategy is mainly applying the prediction technology to predict the traffic amount of immediate reservation traffic in the datacenter interconnected by elastic optical network,and determines whether to remove the idle lightpaths according to the prediction results.However,with the increase of the proportion of advance reservation traffic in the datacenter,the energy saving strategy using traffic prediction for immediate reservation traffic is no longer applicable.For this purpose,applying the software defined networking technology,a batch processing with power consumption consideration strategy is proposed for the advance reservation traffic.That is,traffic combinations are made for different types of advance reservation traffic firstly.Then,the traffic in the same combination is packaged according to the capacity of optical transponder,and the corresponding lightpath is established subsequently.Moreover,the energy consumption of the control link is also taken into consideration.Specifically,to reduce the high energy consumption caused by a large number of idle control links in the network,the currently unused idle control link will be placed in the sleep mode.The simulation results show that the proposed strategy can reduce the energy consumption by about 30% of the network,and it will not increase the bandwidth blocking ratio.