Ericsson: “To achieve the long term goal of 4G and 5G in a Virtualized RAN, we first must create a unified packet network for all the RAN transport interfaces.”
Introduction
2019 saw mobile operators across multiple continents launch 5G networks giving customers a glimpse of new experiences. In 2020, those service providers are scaling out 5G deployments and addressing challenges on the cost side as those networks scale.
To understand how packetizing fronthaul might play a role in this effort, it is instructive to understand the current 4G model. Most cellular networks today use a transport architecture know as distributed RAN (D-RAN). There are typically a dozen or more radios on top of a cell tower, connected to baseband units at the foot of the tower. Each radio has a dedicated fiber optic cable carrying a constant bit rate protocol called CPRI (common public radio interface) from the radio to the basebands unit. The basebands units are connected to a cell site router (CSR) and the data is transported over a backhaul facility to the core network. Backhaul facilities for 4G LTE cell sites are packet based, ethernet connections, and typically require less than 1 Gigabit per second. They are usually fiber facilities but can also be microwave links.
In the case of LTE small cells on utility poles, the baseband units are centralized in a hub site because it is impractical to mount the baseband units at the pole. Small cells only have a few radios and may only require two or three CPRI links between the small cell and the hub site. When CPRI is “hauled” over a long distance it is referred to as a fronthaul facility, and the distance is limited to 10 to 15km because of latency.
Once an operator has started building hub sites for small cells, it lays the groundwork for centralizing the macro basebands as well. As illustrated below, baseband units could be removed from the macro sites and relocated to the hub site that serves the small cells. There are some pros and cons with this strategy.
One of the advantages is that by collocating macro basebands and small cell basebands in the same location, they can be more easily connected together, enabling coordination and carrier aggregation. This allows the macro and small cells to coordinate RF spectrum utilization and minimize interference between them. Another benefit of baseband centralization is that it allows basebands to be more efficiently utilized, potentially resulting in fewer basebands. It also saves space at the macro site, reducing power consumption and cost. However, there is one major problem. Transporting up to 18 CPRI links over 10km (as shown in the diagram) would be very expensive.
One approach to reducing the number of fiber strands is to use DWDM (dense wave division multiplexing) to carry each CPRI link on a separate optical wavelength over a single fiber. The challenge with transporting CPRI, is that it is a constant bit rate, that runs at full rate even when no user traffic is present. So, unlike ethernet, CPRI links cannot use statistical multiplexing to share a common data facility.
The role of packet-based fronthaul in 5G
The new 5G radios have much wider bandwidth and therefore require much more fronthaul data. There are also new technologies like AAS (active antenna systems) where a radio can have an antenna array with as much as 64 transmit and 64 receive antennas. Clearly this would generate far too much CPRI for a fronthaul transport facility.
To address this, 3GPP Release 15 introduced a new baseband architecture that splits the baseband unit into separate logical units. By taking some of the processing functions out of the baseband, and moving them into the radio, the fronthaul interface is simplified and can be transported using standard ethernet technology. This is known as the LLS (Lower Layer Split) and the new fronthaul protocol is called eCPRI.
The remaining functions in the baseband are split into two parts: the Distributed Unit (DU) and the Centralized Unit (CU). This division is referred to as the upper layer split. The baseband functionality that used to reside in one monolithic unit, is now disaggregated over three units, namely the RU, DU and CU. The disaggregation of the baseband facilitates two more benefits: the virtualization of the DU/CU into the vDU/vCU, and “pooling gain” where a resource pool of virtual machines can serve many cell sites. This split architecture is much more conducive to centralization, with the vDU and vCU’s centralized in a datacenter environment, and the more efficient packet fronthaul (eCPRI) connecting to the radios. But what do we do with 4G LTE installed base?
Tens of thousands of 4G LTE radios will remain in the network, and 5G NR radios will be added on existing cell sites. Furthermore, with features like DSS (dynamic spectrum sharing) and mixed mode, some of the 4G LTE radios will be software upgraded so they can simultaneously handle 4G and 5G users. In effect we are overlaying a C-RAN optimized 5G technology, onto an existing D-RAN optimized 4G network. The question now becomes, should we unify the architecture so that 4G and 5G share a common C-RAN architecture?
Centralizing the 4G LTE Network
As mentioned above, centralizing 4G basebands is expensive because of the need to transport multiple constant bit rate CPRI interfaces, one for each radio. But what if we could convert these legacy CPRI interfaces to ethernet?
There are two methods to “packetize fronthaul” to allow CPRI to be carried over ethernet. The first method is called Radio Over Ethernet (RoE IEEE 1914.3). This approach segments the CPRI stream into blocks, places them into ethernet frames, then reassembles the CPRI stream at the far end. This approach doesn’t save a lot of bandwidth, because data is still streaming constantly even when no users are active.
The other approach is to convert CPRI to eCPRI. This is achieved by implementing the “lower layer” baseband functions (as described in the lower layer split above) into a remote device at the cell site. The CPRI streams from the LTE radios are processed and aggregated into an eCPRI stream, which requires substantially less bandwidth. The 5G eCPRI from the new radios, and the 4G eCPRI from the eCPRI converter can be merged into one big ethernet pipe, typically 100 Gigabit ethernet.
Transitioning from C-RAN to V-RAN
The transition plan to introduce new 5G NR radios, and centralize 4G is quite complex and may not be applicable everywhere in the network. For example, rural and suburban cell sites may remain in D-RAN mode, while dense urban areas adopt full centralization. “The new basebands will support both 4G LTE and 5G NR and over time will migrate to a V-RAN configuration where they will run on virtual machines (vDU/vCU).” Ericsson’s Kevin Murphy, technical solutions manager, told RCR Wireless News.
From a transport perspective, we see the emergence of a new type of switch/router that must transport very high data rates (100 GE) and many different time sensitive protocols: CPRI, RoE, eCPRI, Backhaul, and PtP (precision timing protocol) for timing and sync.
“By packetizing fronthaul we can integrate thousands of CPRI interfaces into a unified layer 2 transport infrastructure which can leverage well established ethernet OAM (Operation Administration and Maintenance) capabilities including switching and statistical multiplexing.” Murphy noted.
In summary, implementing packet fronthaul is a critical first step to enable legacy 4G LTE to merge with 5G and evolve to V-RAN.
For more information on 5G fronthaul and other elements of 5G transport, check out the following resource library:
- Building efficient fronthaul networks using packet technologies
- Transport networks have to evolve as 5G scales
- Ericsson prioritizing mobile transport for future cellular networks
- From security to slicing, cell site routers are key to 5G success
- Video: Is microwave ready for 5G?
- Overcoming deployment challenges in 5G networks
- How is 5G impacting the mobile backhaul network?
The post How packetizing fronthaul helps 5G to scale, and decreases TCO appeared first on RCR Wireless News.