Building efficient fronthaul networks using packet technologies

Evolving the transport network is an important first step when deploying 5G on top of LTE.  DRAN has been the dominant architecture for 4G LTE. DRAN will also be a commonly used architecture in 5G NR deployments, with the benefit of reusing the legacy infrastructure investments. But in 5G we will see more centralized deployments especially in dense urban locations. With a Centralized RAN deployment, the baseband units located in a central site and the radio equipment located at the antenna sites are interconnected with a transport network denominated fronthaul, using either the TDM-based Common Public Radio Interface (CPRI) or the packed-based evolved CPRI (eCPRI).

The perceived benefits of CRAN include:

• Reducing OPEX by collecting all baseband units in a central location
• Better utilization of latency sensitive co-ordination services by having intra-site traffic coordination
• Potential pooling and aggregation gains with future RAN evolution
• Better support of vRAN with the ability to build large compute clusters in a more central location
• Improved network management and capacity scalability by having baseband at a centralized location combined with a switched connectivity to add new basebands without re-cabling on remote sites

The value of these benefits should be compared to the additional cost of building, operating, and maintaining the supporting Fronthaul networks that need to meet the stringent demands required by extending the 4G/5G radios to baseband connection over distances of up to 10-15 km.  Ericsson has in close cooperation with operators been designing and building fronthaul networks. Here are six main challenges we’ve discovered when building fronthaul networks:

1. Capacity planning: the capacity demand in a fronthaul network can be as high as 20x higher compared to the backhaul connectivity. Let´s look at an example with a 5G NR Macro antenna site with a capacity in the range of 150-200 Gbps: a typical fronthaul network can aggregate up to 30 macro sites, giving traffic volumes of up-to 6 Tbps in a fronthaul network. It is imperative to have aggregation gains as far out in the network as possible to reduce traffic volumes and number of fibers required. The capacity needs be a challenge if fiber is rented since the cost scales with capacity and characteristics

2. Latency budget: the one-way latency requirement for a Fronthaul network is 75µs, from the baseband to the radio. This latency budget includes the distance and all connectivity devices. A fronthaul network with a 10 km fiber between the hub and the antenna site, would need approximately 50 µs to handle the distance, leaving only 25 µs of latency for the connection devices in the fronthaul network.

3. Timing and sync distribution: in a RAN network, the baseband unit is the sync source, synchronizing the attached radio units. When building a CRAN architecture, the distance between the baseband and the attached radio units increases. The distance itself will have a degrading effect on the sync, but most of the degradation will be introduced by the connectivity devices (optical and packet) in the fronthaul network. When building a fronthaul network, it is important to plan sync distribution when designing the network and to utilize packet connectivity devices with high sync accuracy, but also to limit the number of optical connection devices such as patch panels etc.

4. Operation & maintenance: of a vast point-to-point fiber network without traffic engineering and a range of SFPs at scale can be a daunting task.  To send technical personnel to a remote radio site for troubleshooting and to do manual re-cabling on the hub site, is time consuming and costly.

5. Reliability: when designing the fronthaul network, there is a need to avoid creating a single point of failure that can impact a high amount of macro sites. It is important to understand the traffic flows and to implement redundancy schemes where applicable. For example, if a single transport equipment is used to aggregate the 30 macro sites, failure or maintenance of this transport equipment can impact a large operator serving area.

6. On-going RAN evolution: RAN is evolving rapidly in the 5G space, introducing new features such as spectrum sharing, lower level split and virtualization. The challenge is to evolve the fronthaul network in accordance, supporting new and changing requirements for availability, performance, RAN interworking, connectivity and observability.   

Each fronthaul network will have its own mix of challenges depending on deployment strategy, geographical area, number of sites, and RAN technology.  It is important to evaluate each deployment scenario and apply the right mix of technologies and solutions.  Operators have several technologies that can be applied for the fronthaul use case, each with its advantages and disadvantages. 

These options are described below:

Optical WDM-based solutions

With CRAN, optical technologies are the current preferred solution for fronthaul, including a combination of CWDM and DWDM solutions. These WDM-based options use different wavelengths to aggregate CPRI flows over a single fiber. A wide range of WDM-based options exist today, from pure passive solutions to active components. The advantages of WDM-based optical solutions are:

• High capacity over a single fiber, especially in situations where operators’ have fiber exhaustion or must lease dark fibers
• Transparency from both traffic and synchronization perspectives
• No added latency or jitter as the transport is done by the physical layer (wave lengths)

The main challenge with WDM-based options are the cost of colored optics and limited tools for automation when deployed at scale.

PON-Based Solutions

PON-based solutions, ranging from TDM based to WDM based, can be an option for fronthaul. While PON-based technologies may be used for backhauling, the main challenge for the TDM-based PON systems with traffic engineering (TE) for fronthaul use case, is the introduced latency. To get a 10 km reach of the fronthaul network with, the latency for the infrastructure needs to be below 25 µs but there are no PON solutions (except pure WDM-PON) that can fulfill those tough requirements today.

Packet-based Fronthaul

With the introduction of 5G and new functional splits between the based and radio, there is now a possibility to build a packet-based fronthaul system leveraging the technology we all know and love … Ethernet.  Today there are two main technologies to packetize CPRI:

• Mapping the CPRI frames into ethernet packets, called Radio over Ethernet (RoE)
• Convert the CPRI streams into eCPRI by baseband processing
  
  

RoE mapping can be done in the transport domain because CPRI data is not processed, only transported. The upside to RoE mapping is that it can be done using standard Ethernet technologies.  However, RoE is considered a lowest common denominator approach.  Even though the mapping is done in the transport domain, it still requires knowledge of the RAN as CPRI implementations are RAN vendor specific. The main drawback with RoE is that there is minimal bandwidth reduction in comparison to native CPRI traffic. CPRI is built on the assumption that upstream and downstream traffic flow have the same latency.  To achieve this symmetry, the RoE systems needs to have jitter buffers to handle variations in the fronthaul network. The jitter buffers introduce additional latency, reducing the potential distance of the fronthaul network. Since RoE is built around static CPRI streams, it limits the ability to aggregate and automate traffic flows. You can use Ethernet to avoid the cost of WDM-optics, but since it’s a book-ended solution, not much else is improved in terms of reduced capacity requirement and CAPEX reduction.

The most efficient Ethernet-based fronthaul option is to convert CPRI into eCPRI in the RAN domain.  This approach leverages baseband processing of their CPRI data streams and convert the time domain signal to frequency domain. The benefit of the CPRI to eCPRI conversion is that it drastically reduces the capacity required for ethernet fronthaul by scaling traffic with used antenna bandwidth and by removing the constant bit rate of the CPRI traffic. Since the conversion is done in the RAN domain, the conversion process will not introduce any additional latency in the fronthaul network, enabling longer macro site to CRAN hub site distanced compare to RoE and greater flexibility in building the fronthaul network.

Combining CPRI to eCPRI conversion with packet aggregation at the antenna site is the most efficient approach. The fronthaul capacity demands is reduced by 60%-80% depending on radio configuration, in comparison to other technologies such as traditional CPRI or mapping of CPRI utilizing RoE.

To deliver on this purpose, Ericsson has introduced the Router 6673 for packet-based fronthaul, with embedded RAN Compute functionality. This is enabling efficient conversion to eCPRI for Ericsson RAN sites and also via RoE to cater for older types of radios in the network.

With the Router 6673, operators can use Ethernet connectivity to efficiently build fronthaul networks, and with the same platform deliver carrier ethernet services to enterprise networks and serve as a fixed broadband access aggregation node.

Ericsson has extensive experience building RAN and transport networks and can combine RAN and Transport knowledge to build optimal end-to-end solutions. 

Read more from our whitepaper Packet fronthaul – design choices towards versatile RAN deployments