Efficient proactive caching in storage constrained 5G small cells
Samir Kumar Mishra, Puneet Pandey, Prince Arya, Alok Jain · 2018
With the increase in demand for multimedia data, content caching at the edge nodes like base station (eNodeB) or user equipment (UE) has become an essential requirement in order to meet lower latency for faster access to data in the fifth generation (5G) wireless systems. However, the content caching has many challenges to be solved. The major ones among them are limited storage capability at the edge nodes, prediction of the user preferences and the content update over time. In this paper, we are addressing two use case scenarios, one where caching is performed at the base station and second, where caching is done at the UE which communicates based on the device to device (D2D) communication methodology. For both the use cases, we are addressing two fundamental problems. Firstly, what is the minimum cache size which is needed to attain a defined satisfaction quotient of the network? Secondly, what is the best caching mechanism to attain a minimum satisfaction quotient per user? In this work, we present a novel algorithm for the purpose of popularity estimation of content, named as Rank-Directed Sparse Bayesian Learning (RD-SBL). Further based on the estimated popularity matrix we propose another novel algorithm for content placement at the cache. RD-SBL performs better than the existing solution by providing 30% to 45% per user cache hit to 98.9% users along with better system level cache hit. In a D2D environment as well, RD-SBL provided a significant gain in terms of reduction in memory size and an increase in cache-hit.