Non-uniform EWMA-PCA based cache size allocation scheme in Named Data Networks

TL;DR

This paper proposes a novel cache size allocation scheme for Named Data Networking using a non-uniform EWMA-PCA approach, improving cache hit ratios and reducing unsatisfied Interests by analyzing long-term and short-term network parameters.

Contribution

It introduces a non-uniform cache size allocation method based on EWMA and PCA analysis, enhancing cache efficiency in NDN routers.

Findings

Increased cache hit ratio in NDN routers.

Reduced number of unsatisfied Interests.

Improved cache management efficiency.

Abstract

As a data-centric cache-enabled architecture, Named Data Networking (NDN) is considered to be an appropriate alternative to the current host-centric IP-based Internet infrastructure. Leveraging in-network caching, name-based routing, and receiver-driven sessions, NDN can greatly enhance the way Internet resources are being used. A critical issue in NDN is the procedure of cache allocation and management. Our main contribution in this research is the analysis of memory requirements to allocate suitable Content-Store size to NDN routers, with respect to combined impacts of long-term centrality-based metric and Exponential Weighted Moving Average (EWMA) of short-term parameters such as users behaviors and outgoing traffic. To determine correlations in such large data sets, data mining methods can prove valuable to researchers. In this paper, we apply a data-fusion approach, namely Principal Component Analysis (PCA), to discover relations from short- and long-term parameters of the router. The output of PCA, exploited to mine out raw data sets, is used to allocate a proper cache size to the router. Evaluation results show an increase in the hit ratio of Content-Stores in sources, and NDN routers. Moreover, for the proposed cache size allocation scheme, the number of unsatisfied and pending Interests in NDN routers is smaller than the Degree-Centrality cache size scheme.