NBCC: Simulation of a new Caching strategy using Naive Bayes Classifier in NDN

Abstract

Named Data Networking (NDN) is attracting increasing attention from researchers and companies due to its characteristics and its promised results as a better alternative to the current TCP/IP Internet. Among these features are the use of names instead of addresses, and the use of caches in the nodes. Both have proved to be an excellent addition to network functionality that allow receiving information from a nearby location while relieving the pressure on the main servers. Yet, caches are still limited compared to the huge amount of data consumed. Most research has focused on finding and caching the most relevant data, to retrieve it in the future from the nearest point. Most research agrees that the data that needs to be stored is the one that is constantly requested by many consumers, and this theory has been generally effective in most research works. However, high data consumption levels are not always considered important, especially in academic or corporate environment. This is particularly true whenever the consumption of data associated to the institution’s own servers is very low compared to the other data, such as entertainment videos and private messages from social networking sites. Hence, the data that is stored and delivered in a short time is not essential for these institutions. Also, the servers that are discharged from the pressure are not affiliated with these institutions either. The existence of these cases proved by our study on real consumption data belonging to the Amar Telidji University of Laghouat in Algeria, where we found through simulations that only 4% of the overall traffic is associated with data belonging to the university itself. In this paper, we propose a new placement strategy named NBCC (Naive Bayes Classifier for Caching). The NBCC is used to cache the imported data by classifying the received content using a Multinomial Naive Bayes classifier that can classify the received data using only their names. The strategy is shown to be effective and provides the best results compared to other state-of-art strategies.