MAGIC: A distributed MAx-Gain In-network Caching strategy in information-centric networks

Abstract

Information centric networks (ICNs) allow content objects to be cached within the network, so as to provide efficient data delivery. Existing works on in-network caches mainly focus on minimizing the redundancy of caches to improve the cache hit ratio, which may not lead to significant bandwidth saving. On the other hand, it could result in too frequent caching operations, i.e., cache placement and replacement, causing more power consumption at nodes, which shall be avoided in energy-limited data delivery environments, e.g., wireless networks. In this paper, we propose a distributed caching strategy along the data delivery path, called MAGIC (MAx-Gain In-network Caching). MAGIC aims to reduce bandwidth consumption by jointly considering the content popularity and hop reduction. We also take the cache replacement penalty into account when making cache placement decisions to reduce the number of caching operations. We compare our caching strategy with several state-of-art caching strategies in ICNs. Our results show that the MAGIC strategy can reduce up to 34.50% bandwidth consumption, save up to 17.91% server hit ratio, and reduce up to 38.84% caching operations compared with the existing best caching strategy when cache size is small, which is a significant improvement in wireless networks with limited cache size at each wireless node.

Survey on caching approaches in Information Centric Networking

“…The deployment and location can be an influencing factor for bandwidth utilization, mitigating high hop crosses, minimizing traffic and overall network delay. Studies by Psaras et al (2014), Sourlas et al (2013) and Ren et al (2014) have identified the need for proxy, reverse proxy, active deployments of caches with each related to the localization of the cache-enabled node/router. Therefore, cache size would be efficiently utilized based on the deployment scheme and algorithm of placement, eviction or replacement of contents in ICN.…”

A Comparison of Caching Strategies for Content Centric Networking

“…Especially, a lot of works focus on the design of new caching strategies, as it is a key point of this novel networking architecture [6][7][8][9]. However, these strategies are usually evaluated with dedicated simulation tools, under different simulation environments, scenarios and input parameters.…”

“…127 nodes) where all the requests are issued from the last two levels of the ISP and are served by the root node. [8] and [9] use the same kind of a k-ary tree ranging from 4 to 6 levels and the number of children per node varies between 2 and 5. [10] uses a mixed approach, where an ISP level topology is generated and a binary tree of depth 5 is built at each node of the ISP level topology.…”

“…In [13] [11], α ranges between 0.65 and 2.5. The α varies between 0.7 and 0.96 in [14] while [9] evaluates with an α between 0.7 and 1.1.…”

“…Any nodes having the data can serve future requests, and it helps reducing the load on servers and congestion in the network. Thus, many research works have proposed novel caching strategies in order to improve the performances of CCN, starting from the CCN default strategy Leave Copy Everywhere (LCE) [6], to more sophisticated strategies such as Leave Copy Down (LCD) [6], ProbCache [7], Cache "Less" for More [8] and MAGIC [9].…”