This month, we are focused on virtual routing especially as it related to the new 5G service edge. We recently spoke on an Omdia webinar How to Futureproof your 5G Transport Network. We answered several questions that relate to virtual routing that we thought would be of broad interest.
Question: Today, the 4G RUs and BBUs are co-located at the cell sites with backhaul connectivity from these sites. Going forward, do you see networks evolving to where the exact same 5G BBUs from these cell sites will be relocated to a central site and the backhaul links will be transformed into fronthaul connectivity?
Answer: In 4G, BBUs can be centralized using C-RAN. With 5G, the BBU functions will be implemented as the DU and CU, which allows for even greater centralization and cost savings. Centralization minimizes over-provisioning by pooling resources. Using virtualization, these functions can be implemented at VNFs on commodity hardware.
5G can use any combination of fronthaul (RU – DU communication), midhaul (DU-CU communication), and backhaul (CU-Mobile Core communication). While 5G will not be one size fits all, there are strong financial reasons to use fronthaul. The issues around building out the transport network involve services, latency, and bandwidth. We discuss those in-depth in a new white paper we just published with Mobile World Live.
Q: Do you see virtual routers replacing infrastructure routers from Cisco, Juniper, etc. or do you still see a need for infrastructure routing with some of the edge function moved to virtual routing?
A: There is a massive greenfield buildout of new 5G cell sites that will need routers as part of the transport network. Approaches like TIP’s DCSG call for a router for roughly every three cell sites with the option to connect wireline services on the same network. The greater bandwidth in 5G means existing cell site routers will need to be upgraded to handle both the increased port speeds and aggregate bandwidth. There is significant interest in using open and disaggregated routers because they can meet the cost goals of the MNOs.
Applications like RAN sharing and network slicing will benefit from using virtual routing that is difficult to do on legacy routers. Virtual routers can be fully implemented as software processes running in virtual machines (i.e. VNFs) to run on x86 hardware. However, servers generally cannot meet the throughput requirements as well as an ASIC-based switch. As a result, we expect to see virtual routers implemented on ASIC-based switches in the service edge. Aggregation routers and core routers are obviously also important, so the network will consist of a mixture of these various routers.
Q: If the cell site gateway will be installed in the cell site and carry the fronthaul traffic, won’t the router increase the latency?
A: Yes, but the device latency is negligible.
For example, in the webinar, we discussed timing and synchronization. Precision Timing Protocol (PTP) is covered by ITU-T G.8275.1 (FTS) and G.8275.2 (PTS). Profiles for phase/time distribution are the preferred way to provide time synchronization functionalities in 5G networks, especially in DCSG applications, with all nodes in the network supporting the Boundary Clock (T-BC) role. We have tested our software on ASIC-based switches and the device latency processing and forwarding PTP packets is under 10 nanoseconds. The 10 ns latency puts the router as a Class C device in conformance with G.8273.2 standard. which is the most stringent class. The latency budget to connect a RU to DU on fronthaul is 100 microseconds which means the DU may be up to 10 km away from the RU. The device latency is not an issue for a well-designed fronthaul network.
Q: Will disaggregated hardware/software really work in the fronthaul portion of the 5G transport, considering the stringent latency and synchronization requirements? This may work for backhaul but will not work for fronthaul.
A: There are two different issues in this question. The first relates to the latency and synchronization issues by having a router at or close to the 5G cell sites. This problem is common to all routers. Clearly, they must support the relevant features like PTP. We addressed device latency in the last question. Routers add value when IP transport is used in the fronthaul and by enabling features like MPLS and segment routing for traffic engineering.
The second issue this raises deals with disaggregated routers. Is there something fundamentally different about the performance of the router in a disaggregated implementation? From a performance standpoint, the answer is no. Routing protocols are standardized and well understood. Performance and interoperability can be tested and validated relatively easily in the lab. Moreover, the white-box switches use commercial ASICs that are increasingly being used by traditional router vendors. For example, both Cisco and Juniper use the Broadcom Qumran chip, which has been specified by TIP for the DCSG project.
The disaggregated model offers several advantages including lower cost, less vendor lock-in, and more opportunities to innovate.
For a deeper discussion on routing and 5G transport, please get the white paper, The Future is Virtual: Routing in 5G Transport Networks, on GSMA’s Mobile World Live.