Entropy-based probabilistic cache replacement strategy in named data network

Abstract

In the named data network (NDN), the typical LRU and FIFO cache replacement strategy only considers the time and progress of a single influencing factor, and cannot distinguish whether the frequency of content requests is high or low, and there is a problem that popular content is expelled by non-popular content. In order to achieve efficient NDN cache replacement, an entropy-based probabilistic cache replacement strategy (EPR) was proposed. This strategy was expanded on the basis of the original format of the data packet, adding 3 fields to record the size of the cached content, content popularity and request cost, and making each routing node count the data carried by all the data packets of the node when it needs to replace the data packet. Based on the information of these three fields, the entropy weight value and replacement probability of each data packet were calculated according to the attribute value and the assigned attribute weight, and finally the cache content was replaced based on the calculated replacement probability. Experimental results show that compared with some common NDN cache replacement strategies, this strategy can effectively increase the average cache hit rate and reduce the average request latency.

Keywords

• named data network
• cache replacement
• information entropy
• cache hit rate