Mobile Operators need to carefully design their networks and 5G gNB deployment options according to the existing network architectures and the service demands when it comes down to operational costs and ease of deployment.
3GPP has specified a new interface within the 5G radio node’s baseband. This new interface allows to split the 5G gNB’s baseband hardware and to place parts of baseband functionalities in more centralized units. With that approach, cooperation between the neighboring nodes, previously requiring a cell site to cell site communication, can now be handled within the baseband unit’s centralized part.
5G gNB Deployment challenges
Whereas the higher layer split (HLS/F1-interface) has nearly the same transport requirements as the S1 interface, the low layer split (e-CPRI or CPRI interface) imposes strict requirements with regards to bandwidth and latency towards the transport network.
Regarding 5G latency requirements, one-way transport delay should not exceed 100ns for the eCPRI/CPRI interface. According to radio vendors’ feedback, the eCPRI interface, even if it uses Ethernet on Layer 2, cannot be aggregated via an Ethernet switch or IP router due to the strict latency and jitter requirements. This is because it is not supported by any switch or router product currently available on the market. Consequently, dedicated fiber/wavelength for each radio unit would have to be considered even for eCPRI.
One option to reduce the number of fiber/wavelength by a factor of 1 to 3 is chaining of the RRH (CPRI chaining). This is an interesting solution, however, SPFs are potentially more expensive. It is also worth mentioning that the amount of fiber links per wavelength above does not consider the centralized geo-redundancy of the baseband. In case geo-redundancy is implemented, the amount of fiber links per wavelength is twice as high.
Such a deployment scenario had already been calculated for LTE. However, even without geo-redundancy, this deployment scenario is not that cost-effective.
Therefore, lower layer split is not easily applicable for network-wide deployment. Only in exceptional cases like specific industrial areas or campus network solutions can lower layer split be deployed, at least in the early 5G gNB deployment phase. Therefore, there is no immediate need to consider a lower layer split in the transport design. Furthermore, as the higher layer split (F1-interface) is from a transport requirement perspective comparable to the classical deployment (S1 interface), it also does not need to be considered currently in the transport design.
5G Early deployment
The early 5G deployment focuses on the traditional deployment approach. Nowadays, most of the 4G cell site connections are via 1GE fiber or Microwave links. When adding a 5G gNB to a 4G site, it is preferred to upgrade the cell site’s connectivity to at least 10GE.
During the early deployment phase, it is neither recommended to use multi-standard base stations (currently not even offered by any vendor) or use the 5G gNB for traffic aggregation. The 5G software will not be mature in the initial phase, and several software upgrades and patches are expected for optimum performance. The cell site’s implementation should ensure that 4G is unharmed if 5G software is patched or upgraded.
Nowadays, even though network upgrades are ongoing as we speak, most existing eNBs are only equipped with 1GE interfaces. In these cases, it would not be possible to route the 5G traffic via the 4G eNB. That would be an option only in cases where 4G eNBs have already been upgraded to 10GE.
If geo-redundant connection towards redundant backhaul routers is also considered, the most standardized solution would be to introduce a dedicated Cell Site Gateway to aggregate the traffic of the eNB and the 5G gNB to a common transport link. That Cell Site Gateway (CSG) can also be used to implemented a radio vendor-independent, harmonized geo-redundancy solution.
This is also applicable if the cell site is also used as a microwave hub site. The CSG can also aggregate the traffic coming from the MW tree onto the common transport link. It can provide the same geo-redundancy for the radio node connected via MW and route X2/Xn interfaces between local and via MW connected radio nodes.
X2/Xn interface
The expectation is that the Xn interface between the 4G eNB and the 5G gNB can utilize the standard transport interfaces. That means the Xn-interface between the collocated nodes should be routed via the CSG. Today not all vendors do support that option. Some vendors use a proprietary back-2-back link between the two radio nodes. This is not a very flexible solution, as that proprietary interface forces us 5G gNB from the same vendor as the 4G eNB is.
Another approach is the 5G small cell deployment. Small Cells are also expected to follow the non-standalone (NSA) approach. However, these small cells might still use a 4G macro site as an anchor layer. In that case, a back-2-back link is not possible or only possible with extra effort. This complexity contradicts the expectation of a simple and cost-efficient 5G small cell deployment.
To improve the gain of cooperation features between radio nodes, it is recommended to minimize as much as possible the complexity. This means on the one side to route the traffic as close to the cell site as possible, and for collocated radio nodes or MW trees to route X2/Xn directly on the CSG and/or the MW tree.
This already implies routing capabilities on the CSG, as a L2 solution at the cell site gateway is not manageable for regular networks. That would require re-planning of IP addresses and planning site-specific subnet sizes, which is a clear disadvantage compared to the current standardized site solution.
Dynamic vs Static routing
Therefore, L3 routing functionality could be implemented at the CSGs. It would also be wise to use dynamic routing protocols between the CSG and the mobile backhaul. For example, in a case where cell sites are geo-redundantly connected towards the mobile backhaul, CSGs may have a short fiber link as primary towards the nearest metro ethernet router and a long fiber link to a backup metro router. For regular traffic towards the Core Network, the primary path of each CSG would be the best choice. In this case, static routing would be sufficient. But for X2/Xn, static routing would mean routing deep in the mobile backhaul, with the corresponding delay impact. Using dynamic routing at the CSGs, the shortest path via the common backup metro router would be selected, and overall latency could be improved.
