Resource management with adaptive capacity in C-RAN

Abstract

Efficient computational resource management in 5G Cloud Radio Access Network (CRAN) environments is a challenging problem because it has to account simultaneously for throughput, latency, power efficiency, and optimization tradeoffs. This work proposes the use of a modified and improved version of the realistic Vienna Scenario that was defined in COST action IC1004, to test two different scale C-RAN deployments. First, a large-scale analysis with 628 Macro-cells (Mcells) and 221 Small-cells (Scells) is used to test different algorithms oriented to optimize the network deployment by minimizing delays, balancing the load among the Base Band Unit (BBU) pools, or clustering the Remote Radio Heads (RRH) efficiently to maximize the multiplexing gain. After planning, real-time resource allocation strategies with Quality of Service (QoS) constraints should be optimized as well. To do so, a realistic small-scale scenario for the metropolitan area is defined by modeling the individual time-variant traffic patterns of 7000 users (UEs) connected to different services. The distribution of resources among UEs and BBUs is optimized by algorithms, based on a realistic calculation of the UEs Signal to Interference and Noise Ratios (SINRs), that account for the required computational capacity per cell, the QoS constraints and the service priorities. However, the assumption of a fixed computational capacity at the BBU pools may result in underutilized or oversubscribed resources, thus affecting the overall QoS. As resources are virtualized at the BBU pools, they could be dynamically instantiated according to the required computational capacity (RCC). For this reason, a new strategy for Dynamic Resource Management with Adaptive Computational capacity (DRM-AC) using machine learning (ML) techniques is proposed. Three ML algorithms have been tested to select the best predicting approach: support vector machine (SVM), time-delay neural network (TDNN), and long short-term memory (LSTM). DRM-AC reduces the average of unused resources by 96 %, but there is still QoS degradation when RCC is higher than the predicted computational capacity (PCC). For this reason, two new strategies are proposed and tested: DRM-AC with pre-filtering (DRM-AC-PF) and DRM-AC with error shifting (DRM-AC-ES), reducing the average of unsatisfied resources by 99.9 % and 98 % compared to the DRM-AC, respectively.