Strategies for Load Balancing in Multi-tenant 5G C-RAN Architecture

Abstract

The availability of resources to meet the ever-growing demand of traffic limits the efficiency of existing cellular networks. The next-generation cellular platforms need to be scalable, flexible and support multiple radio access technology (M-RAT) to overcome the resource availability problem. Moreover, to minimize capital expenditure (CAPEX) and operational expenditure (OPEX) there is also a need to modify the network architecture and control strategy. The 5G C-RAN architecture is in this direction to meet the above objectives. Centralized processing and controlling in C-RAN reduce associated cost and make the base station energy efficient. However, centralization of all baseband processing increases computational complexity and associated delay. An unbalanced condition at remote radio head (RRH) and centralized base band unit (C-BBU) arises when utilization of physical resources goes beyond a certain threshold limit due to an increase in traffic or processing load. Due to unbalanced conditions, the system performance at RRH and C-BBU worsens and increases corresponding power consumption. This thesis has attempted to enhance resource utilization by distributing the load across the available processor at C-BBU. A co-operative load balancing (CLB) technique is proposed to maximize resource utilization and system throughput by sharing traffic and processing load between RRHs and C-BBUs. Efficiency of the proposed co-operative approach is evaluated using different parameters like waiting time, blocking probability, and processing time at RRH and C-BBU. It is observed that the CLB has the lowest waiting and processing time as well as the blocking probability compared to the existing contemporary algorithm. For load balancing across the BBU-pool, an inter-BBU migration is proposed; here, VBs from an unbalanced C-BBU is migrated to another C-BBU within the same BBU-pool. Performance results show that migration based load balancing reduces migration time and downtime by a significant amount. To reduce load of fronthaul and C-BBU, a multi-tier delay aware load balancing (MDALB) algorithm is proposed. A load distribution factor is used to estimate available capacity for packets allocation inorder to minimize system delay and packet loss. A power consumption analysis for traditional D-RAN and C-RAN was carried out. The analysis shows that C-RAN provides 40% to 50% energy efficiency compared to D-RAN architecture.