5G and Disaggregated Routing – A Perfect Match?

5G requires a major upgrade of transport infrastructure in order to provide up to 10 Gb/s to users and a 1,000-fold increase in capacity per unit area, as well as lower latency, higher availability, and sliceability.

Multi-access edge computing (MEC) is necessary to meet the low latency requirements of 5G’s ultra-reliable low-latency communications (urLLC), scale for massive machine-type communications (mMTC), and support emerging applications such as augmented reality.  While the scalability of IP routing has long made it the technology of choice for mobile backhaul, it will have an even stronger role to play in 5G as traffic patterns become more meshed and less predictable with the adoption of MEC, as shown in Figure 1.

Figure 1: Meshed 5G traffic patterns require IP routing

Therefore, 5G represents a key use case for disaggregated routing, as evidenced by the Telecom Infra Project (TIP) Disaggregated Cell Site Gateway (DCSG) and Open Compute Project’s Cell Site Gateway Router initiatives, which we discussed in a previous blog, as well as in the results of a recent Infinera-sponsored report, “Operator Strategies for Disaggregation in 5G Transport Networks” by Mobile World Live. In the survey that formed the basis for this report, the majority of respondents indicated that they were “very likely” or” somewhat likely” to include disaggregation in their 5G transport, as shown in Figure 2.

Figure 2: Disaggregation and 5G transport survey question on disaggregated architectures

The question remains: which parts of the 5G transport network can disaggregated routing address? To answer, we need to look at our options for how to split the functionality of the base station, as shown in Figure 3. At a high level, it splits into the radio unit (RU), distributed unit (DU), and centralized unit (CU). However, there are a number of options for how to split and combine these elements.

Figure 3: 5G xHaul options

The RU, DU, and CU can be colocated like in a traditional base station, shown as Option A. Option B combines the RU and DU to create a midhaul network and a backhaul network, while Option C combines the DU and CU to create a fronthaul network and a backhaul network. Option D completely separates the RU, CU, and DU, with fronthaul, midhaul, and backhaul networks. Early non-standalone 5G deployments focused on enhanced mobile broadband (eMBB) are likely to be weighted in favor of Option A, evolving to the more distributed options (B, C, and D) as mobile operators move to standalone flavors of 5G and extend their service portfolio to urLLC and mMTC.

While disaggregated routing can provide a solution for all aspects of 5G xHaul, specific requirements regarding scalability, availability, latency/jitter, and environmental hardening may apply depending on the type of xHaul and location in the network. Environmental hardening is likely to be required for access locations, while multi-unit scaling may be required for the aggregation and core parts of the backhaul network, with stacking in the aggregation nodes and fabric-based scaling in the core.

Regarding latency, while there are some inconsistencies around the exact requirements, as a rule of thumb backhaul requires maximum latency of around 10 ms, while midhaul’s requirements are an order of magnitude lower at 1 ms, and fronthaul’s yet another order of magnitude lower at 100 µs. With merchant network processors able to deliver latencies of under 10 µs, and even under 3 µs, disaggregated routing can provide a good fit for the access, aggregation, and core network parts of a backhaul network, as well as for midhaul.

Fronthaul, however, is where it gets interesting. While the eCPRI protocol used for 5G fronthaul can be transported over Ethernet or UDP/IP, according to the IEEE 802.1CM Time-Sensitive Networking for Fronthaul standard, one-way latency must be less than 100 us. Assuming a distance of 10 km at 5 us per km, that leaves 50 us for transport equipment, while jitter must be less than 5 us or 10% of latency and packet loss must be better than 10^-7.

Of these requirements, jitter is probably the most challenging to meet, and requires a network processor that delivers very flat latency. 802.1Qbu Frame Preemption, which can pause the transmission of a lower-priority packet that has already begun transmission when a higher-priority packet arrives, can help, especially on 10 GbE or 25 GbE interfaces.

For more information on this important topic, see the Infinera white paper “The Case for Disaggregated Routing in 5G and DAA Transport Networks”.