A Dynamic Replication Scheme of User Plane Data over Lossy Backhaul Links Yekta Turk a and Engin Zeydan b a Ericsson Research, Istanbul, Turkey, 34396 b Centre Technologic de Telecomunicacions de Catalunya (CTTC), Castelldefels, Barcelona, Spain, 08860 1 Emails: yekta.turk@ericsson.com, engin.zeydan@cttc.cat Abstract —5G networks are characterized by strict latency, jitter and reliability requirements. On the other hand, a lossy backhaul scenario where high packet losses over the backhaul transmission medium can render backhaul links unavailable for service consumption. This necessities a multi-transmission scheme over backhaul using the existing resources of the opera- tors such as base stations and core networks. This can enhance reliability for 5G backhaul networks to meet the requirements of new 5G services that have strict end-to-end latency re- quirements. In this paper, we propose a GPRS Tunnelling Protocol User Plane (GTP-U) Protocol Data Unit (PDU) multi- transmission scheme that achieves improved Transmission Con- trol Protocol (TCP) latency and packet loss probability at the expense of higher resource (bandwidth) utilization compared to existing traditional systems. The scheme can use several measuring mechanisms such as continuous Two-Way Active Measurement Protocol (TWAMP) testing between core network and base stations, GTP-U user plane echo requests/replies or backhaul quality testing during evolved Packet System Radio Access Bearer (eRAB) establishment phase. This mechanism can also be applied to X2 interface between base stations to reduce the service latency and improve the backhaul link performance. Our simulation results using ns-3 indicate that by applying the proposed strategy in backhaul systems, the application level packet loss and TCP delay can be improved at the expense of the bandwidth improvement at certain drop or loss levels. Index Terms—5G, user plane, GTP-U, backhaul, multi- transmission. I. I NTRODUCTION Time-constraint applications and services can be con- sidered to be the most important feature that are promised with 5G networks [1]. However, there can be several sce- narios where a sudden decrease in the service level of backhaul links in Mobile Network Operator (MNO)’s existing infrastructure can deteriorate the service. These scenarios are predominantly ad-hoc in nature and should be solved using the existing infrastructure facilities in short duration. Therefore, it is highly desirable to be able to create a dynamic self-organized reliable solution to backhaul fail- ures in such conditions. Additionally, existing infrastructure should also be utilized with software modifications that can be integrated to already existing infrastructure. Otherwise new deployments of network elements need to be followed. GPRS Tunneling Protocol (GTP) protocol is specific tun- neling protocol used between end nodes in General Packet Radio Service (GPRS) backbone network and is a group of Internet Protocol (IP)-based communications protocols used to carry general packet radio service. It includes both signaling and user data transfer procedures in interfaces such as S1-U, S10, S11, S4, S5/S8 and S3 in cellular network infrastructure. Generally GPRS Tunneling Protocol Control Plane (GTP-C) and GPRS Tunneling Protocol User Plane (GTP-U) protocols are used for control and user planes of Packet Switched (PS) traffic. Each user plane connection requires GTP-U tunnel across the Evolved Packet System (EPS) backbone network to transmit the user payload. Hence, GTP-U is used to transfer the user data in separated tunnels for each Packet Data Protocol (PDP) context. GTP- U can be seen as a framing protocol which allows multi- protocol packets to be tunneled through the EPS backbone to provide a service for carrying user data packets. On the other hand, GTP-C tunnel, which is the control section of the GTP standard, is used to create, modify and delete tunnels. An example for the end-to-end latency requirements of 5G services for vertical industries is summarized in Table I as described in [2]. In an end-to-end access network, new generation 5G networks should be provided with very fast and critical communication to meet the stringent numerical requirements outlined in Table I. Moreover, this fast trans- mission should not only be available in the radio interface, but in the backhaul interface as well. Improvements on NG- UU interface are not adequate to improve the end-to-end latency of communication systems in cases where backhaul network becomes the bottleneck of communication. The slightest error in backhaul networks can yield inefficiency TABLE I 3GPPP DEFINED L ATENCY NEEDS FOR VERTICAL I NDUSTRIES [2] Service E2E Latency Requirements Live Streaming < 20 ms Smart Grid < 50 ms Time-critical sensing < 30 ms Real-time control for discrete automation ≤ 1 ms Remote drone operation & Cooperative farm machinery ≤ 30 ms Real-time video < 100 ms