Elastic-RAN: An adaptable multi-level elasticity model for Cloud Radio Access Networks

Abstract

Cellular mobile networks in 2020 will increase ten times their coverage area, with more than 50 billion connected devices. We will lead to a massive increase in data traffic, also fostering the development of 5G networks. Therefore industry and scientific initiatives have a crucial role in proposing related projects to meet such demand. Cloud Radio Access Networks (C-RANs) are gaining more and more attention in this context by adopting an architecture in which baseband units (BBUs) run into cloud computing resources, therefore taking advantage of distributed systems flexibility and cloud elasticity. One of the significant challenges in C-RANs lies in the high complexity of orchestrating computational resources to process incoming requests with both high performance and low infrastructure cost. In this regard, this article presents the Elastic-RAN model, which proposes a multi-level and adaptable elasticity for C-RANs. First, we explore the multi-level feature as follows: (i) one level for the BBU pools (i.e., physical machines), given the high volume of traffic to particular BBU pools; (ii) another level for BBUs themselves (virtual machines) due to the high CPU and memory demands to process the incoming requests. Second, the adaptive feature refers to the moldable elasticity grain which resources in both previous levels are provisioned as close as possible to the current processing needs. We evaluated Elastic-RAN through experiments that simulated different load profiles, considering both CPU and network demands. We observed that Elastic-RAN might achieve gains up to 64% in the execution time when compared to a traditional C-RAN. Cellular network operators using the proposed technique will spend less energy and will have a solution that dynamically adjusts the baseband signal processing accordingly to the demand in their access networks.