5G technology is about to deliver a large variety of exciting but also demanding use cases to the end-user. From super-fast speeds for 5G enhanced Mobile Broadband (eMMB) services to massive 5G connectivity for the support of Massive Machine type communications (MMTC), 5G networks will also need to address strict latency and reliability requirements to cover the Ultra-reliable, Low-latency Communications (URLLC) critical applications.
Mobile Operators need to find a way to develop a flexible and scalable 5G network that will be able to offer resilience to support all types of applications with different service requirements. This new 5G network architecture should be able to create virtual software-configurable routes that can be rapidly reserved or dynamically appointed on demand according to each specific service needs.
Existing network architectures
Most Service Providers operate their existing networks in a hybrid L2 and L3 network architecture. An eNodeB is connected to either a microwave or metro ethernet switch and the service is initially backhauled in a L2 fashion. L2 Traffic is mapped to specific customer C-VLAN IDs and is forwarded via E-Line and E-Lan services, without the need for IP routing. The bandwidth is shared among different services with specific bandwidth profiles and service instances.
On the aggregation node, a transport/router device typically forms an MPLS tunnel, and the traffic is now transported over the MPLS domain. In IP/MPLS or L3VPN technology, traffic forwarding is L3 based. Traffic is forwarded over the MPLS tunnel via pseudowire (PW) services. The service tunnel is formed by creating virtual circuits known as Label Switch Paths (LSPs) across the MPLS network. MPLS supporting devices should make up the MPLS network infrastructure.
5G Network Architecture
5G networks can take advantage of the latest network architecture techniques such as Network Slicing, Network Function Virtualization (NFV) and Software Defined Networks (SDN) to deliver state-of-the-art end-to-end services.
Using Network Slicing, 5G networks can form virtual routes per specific service, according to the service needs. NFV can assist in creating virtual blocks to provide customized paths, and the SDN controller can then route the services over the appropriate blocks in a scalable and software configurable manner by separating the user from the control plane. 5G network architecture can, therefore, provide advanced routing techniques and guarantee a flexible and enhanced end-to-end network performance.
Multiprotocol Label Switching (MPLS)
In an MPLS environment, the packets are forwarded by encapsulating them in an MPLS header containing one or more MPLS labels. Label Switching Routers (LSRs) are MPLS capable devices that perform MPLS encapsulation. The ingress LSR is the first router on an MPLS domain that receives the packets and prefixes the MPLS header. The ingress LSR consults the Virtual Routing & Forwarding (VRF) table to decide which MPLS header has to be attached.
The data is then transported over MPLS label-switched paths (LSPs) which are typically set up via LDP and RSVP-TE protocols. OSPF and IS-IS can then be used for the VRFs. The packets are routed according to their labels across the label-switched path. The routers along the path make the forwarding decision based only on the MPLS header. The MPLS headers are removed on the egress router (egress LSR), and the packets are delivered in their original format.
MPLS segment routing
MPLS segment routing provides a simplified routing procedure compared to standard MPLS. LDP and RSVP are no longer required, since, with MPLS segment routing, a label stack is defined at the ingress router that determines the whole path through the network. In an MPLS segment routing domain, the segment is represented by a label and a series of segments by a label stack. This way, traffic is steered through the network path without the need for extra intelligence and signaling.
Traffic differentiation can be achieved for example by forwarding URLLC services and eMMB services via a different network route. Every different service flow can have a dedicated label stack to lead traffic via the MPLS network path according to its specific policy.
Segment routing could also be extended up to the cell site, especially in the case of geo-redundant local loop for URLLC services. MPLS Segment Routing uses the MPLS data plane, where the segment is identified by the MPLS label. Most networks that operate in IPv4 can take advantage of the existing MPLS technology that is already supported by most devices.
IPv6 segment routing
A segment can be typically represented via MPLS labels or IPv6 extension headers. In IPv6 Segment Routing (SRv6) the network uses the IPv6 address to identify the specific segment. When IPv6 is expanded and supported by all network elements, segment routing can be available by using the IPv6 extension header without the need for MPLS support.
Segment Routing and the SDN controller
The SDN controller can provide an end-to-end view on the network chain. It may monitor the network and propose the appropriate changes where needed. The SDN controller will be able to suggest the optimum path across multiple domains by calculating the path with the shortest delay based on latency measurements on the network or by identifying the path with the required, available capacity per service.
Once the SDN controller determines the desired routing path based on these specific service criteria, it is then responsible for providing these optimal routing path instructions to the LSRs.
5G network migration to segment routing
Most existing networks are using IPv4 and are equipped with MPLS enabled or MPLS hardware ready elements. Therefore, migration from existing MPLS networks to MPLS segment routing seems feasible since most of the existing devices would require only a software upgrade, while new deployments could focus on MPLS capable elements.
Segment routing in the mobile backhaul could also extend up to the cell site for higher flexibility. L3 site aggregators can be used to create a L3 end-to-end communication with scalable and dynamic routing capabilities. By supporting a L3 end-to-end path and by placing the application within the radio node, direct connection between the radio nodes can be established, thus reducing the communication path and the corresponding delay times to support ultra-low latency applications.
Operators could also focus on supporting IPv4/IPv6 dual stack. IPv4 could interwork with existing infrastructure and support the early 5G deployments while IPv6 could ensure future proof operation in a standalone 5G Core Network with millions of connected devices and various demanding service applications.
Mobile Operators can take advantage of the above techniques to develop flexible 5G networks and fulfill future 5G requirements without considerable effort and at a reasonable cost.
