The Effect of Cache Partitioning and Sharing on Named Data Network

Abstract

The IP network is inefficient to support data communication. The IP network modifications into a Content Distribution Network scheme can not accommodate consumer mobility and consumer demand patterns that is change dynamically. So that, Named Data Network has been proposed to focus on content for data communications. The Named Data Network router comes with a content store for storing data. The content store size is limited so special techniques are required to optimize content store usage and keep NDN performance in good level. In the previous study, the majority choose data to be stored in content stores based on content popularity only. Popular data will be prioritized to keep in content store and the data with less popularity will be deleted from content store when it is full. This basic technique is known as ‘sharing-technique’. Sharing-technique is difficult to differentiate the treatment of the data with more than one classification requirement. In fact, it is needed to differentiate the treatment for different requirements classes, not only the content priority. In this research, simulation is conducted to analyze the influence of content storage policy with sharing techniques and partitioning-techniques. The partitioning- technique accommodates data storage that considers to the different class of content and also popularity of the content. The results of the simulation show that partitioning-technique gives maximum 36% more cache hit ratio than sharing-technique, viewed in network level. Partitioning-technique allows flexibility of class-based settings for content based on the proportion of content store, while in the sharing technique the difference in cache hit rate for content classes is not significant.