Optimized Planning of DU/CU Placement and Flow Routing in 5G Packet Xhaul Networks.

Packet-switched Xhaul networks (PXNs) have been proposed as a cost-effective and scalable solution for provisioning of connectivity between densely located radio antenna (remote) units (RU) and distributed (DU) / central (CU) processing units in 5G Radio Access Networks (RANs). A PXN enables statistical multiplexing of different data flows, such as fronthaul, midhaul, and backhaul flows, and their routing over a commonly shared packet transport network, thus increasing network flexibility and decreasing bandwidth requirements. The decisions concerning the placement of virtualized DU and CU entities at selected processing nodes and the selection of paths for routing of data flows between these nodes have a direct impact on flow latencies. In addition, buffering of packets in switches introduces dynamic, flow-dependent latencies, which as well have to be accounted for in latency-sensitive 5G RANs. In this work, we address a network planning problem that concerns the joint latency-aware placement of DUs/CUs and routing of data flows (LADCPR) for a set of RUs in a 5G RAN connected using a PXN. We model the LADCPR problem as a Mixed-Integer Linear Programming (MILP) optimization problem and propose two reformulations of the model that facilitate its solving. Since the MILPs have limited scalability, we develop two heuristic algorithms, based on problem decomposition and an iterative search approach. Numerical experiments performed in different network scenarios show that the algorithms are capable of generating good-quality solutions to the LADCPR problem. Also, they can optimize larger network instances – consisting of tens of switches, a number of routing paths, and comprising some hundreds of demands – within relatively low algorithm run times.