In recent years, as Internet services have diversified and grown rapidly, intelligent network environment has dynamically changed with parameters such as traffic patterns and network topologies. To flexibly manage such dynamic changes, SDN (Software Defined Network) technology has emerged. SDN enables more flexible Internet services by dividing the network architecture into a control plane and data plane. By the way, in the SDN, a problem with flow entry replacement may arise owing to flow table size restrictions within the switches. A flow entry replacement problem can increase the packet processing time and degrade the quality of service for the users. Therefore, we need to know exactly the performance of the flow entry replacement algorithms. In order to practically analyze the performance of the flow entry replacement algorithms, we first collect and analyze the actual Internet traffics of famous Internet services such as Instagram, Facebook, Youtube, and Netflex. Then, we analyze the performances of flow entry replacement algorithms by the collected traffic data. Based on the results, the LFU (Least Frequently Used) algorithm exhibits the worst performance, whereas the FIFO (First In First Out), LRU (Least Recently Used), and SFF (Short Flow First) algorithms show relatively better performances.
展开▼