Mobile Backhaul

PDF versie

Mobile operators are looking to develop fresh service portfolios to differentiate their offering from their competitors, retain customers and increase use of data services. Now, successful broadband mobile data services must be useful and entertaining. And, more and more, end-users are demanding consistency, quality and reasonable prices.
To cope with increasing demand for broadband mobile data services requires operators to upgrade their 3G networks. Furthermore, using new wireless technologies such as WiMAX and LTE for broadband connectivity requires a dramatic increase in capacity for cell sites. As a result, pressure mounts on the transport network.

Backhaul network implementation must be designed to support multiple mobile operators sharing the same cell tower. Because cell tower sharing is a common practice in most countries today, it has become a business imperative that the fixed network operator must plan to serve multiple mobile operators from the same backhaul network. According to industry estimates, more than 60 percent of all cell towers worldwide are shared by multiple mobile operators.

Understanding Mobile Backhaul

Mobile backhaul encompasses the transport network between cell sites (base stations) and associated controller or gateway sites. Cell sites are numerous, often geographically remote and, up til now, with limited bandwidth.
Conventionally, the BSC (Base Station Controller) and RNC (Radio Network Controller) mark the border between the access network and the core network. However, this border is being blurred by new radio technologies, like LTE, and WiMAX, which introduce flat architectures, in which cell sites are connected directly to the core network.
Even though the network is more and move moving from a hierarchical to a more flat architecture, an access network remains necessary as the cell sites are graphicaly dispersed.
 

^{Figure 1: Traditional GSM network}

Wireless networks have historically relied on TDM transport services for interconnection between cell sites and base station controllers (BSC) and mobile switching centers (MSC) . Looking at this, two qestions arise: why do wireless operators rely on TDM services, and why haven’t they shifted to an Ethernet/IP infrastructure?
The wireless industry developed standard interfaces for interconnecting different functional devices within their networks, such as the base station transceiver (BTS) to BSC to MSC. These interfaces go by an alphabet soup of names including the A, Abis, Gb, and Iu interfaces, as shown in Figure 1.
As data services were added to GSM cell sites, new elements were introduced into the network.
An EDGE (Enhanced Data rate for GSM Evolution) blade is typically installed in the base station transceiver node to support data services up to 400 Kb/s. The EDGE blade communicates over the Gb interface, as shown in Figure 1, to the Serving GPRS/EDGE Support Node (SGSN) located in the mobile switch center. Voice services continue over the A interface, while data
services are handled over the Gb interface.

Similarly, 3G networks have their own set of defined interfaces between their base stations, (Node B), radio network controllers (RNC) and voice and data switches, as shown in Figure 2.
 

^{Figure 2: 3G network}

These Abis, A, Gb, and Iu interfaces have historically defined the backhaul transport requirements, since they specify the entire protocol stack, including the physical layer 1 implementation. The wireless industry’s heavy reliance on E1 and T1 transport services is defined by and required by these industry specifications.
The 3G/UMTS standard and it's successors have however specifications for Ethernet also.
The use of Ethernet as transmission protocol for voice will be discussed later as this has some specific requirement as timing is very critical.
 

Radio Access Network Topology

While the conventional, 2G, 2,5G and 3G networks are based on a centralized a.k.a. star network where all cummunication is routed through a Base Station Controller, it's successors, 4G, LTE and WiMAX networks are flat networks where each Base station can have communication with another or with it's gateway. This is also shown in the pictures 3.

^{figure 3: traditional versus 4G network topology}

This next generation RAN will also introduce a new flatter architecture in the core also that will eventually cause the traffic to be
more dynamically distributed in the transport network. This is why any RAN upgrade is intrinsically linked to the evolution of a backhaul network. Backhaul has to evolve from TDM to packet based architecture in order to deliver economically efficient solution.
The issue for operators is to decide the timing and pace of their technology migration in the Radio Access Network. If an operator opts for a conservative approach and slow rollout there is a risk of missing the opportunity and losing competitiveness. On the other hand, the success of a more aggressive rollout will depend on the alignment of radio access and backhaul transport strategies.

Wireless capacity

Given the wireless industry’s historical reliance on E1 and T1 circuits, a good understanding of the actual capacity requirements is critical to any mobile backhaul service. While in the traditional GSM most cell towers could be serviced by only one to four E1 or T1s, equivalent to 1.5 Mbps to 6 Mbps. The addition of 3rd and currently the upcoming 4th generation data services increased the need for more bandwidth to cell sites dramatically.

The amount of bandwidth required at a cell site is constrained by two factors:

• the amount of wireless spectrum available
• spectral efficiency of the wireless interface.

Wireless frequencies, or spectrum, is allocated and auctioned by the government or, in the US, by the FCC, typically in 10- or 20-MHz blocks. Half of each block is used for transmitting signals and the other half for receiving signals. Frequency blocks are further subdivided into “channels” that are shared across cell areas.

Spectral efficiency is the amount of data (bits/second) that can be transmitted for every Hz of spectrum. Technologies, such as EDGE and HSDPA utilize advanced modulation schemes allowing higher data rates by squeezing more bits/s into the allotted amount of spectrum. These advanced modulation schemes dynamically adjust depending on the channel conditions between the base station and handset (power, noise, interference, etc.).
There is a natural upper limit on cell bandwidth that is simply the amount of spectrum owned and available at a cell site multiplied by the spectral efficiency of the wireless interface (bits/s/Hz).

As an example, assume we have a three-sector cell site providing 2G GSM voice services over 1.25 MHz of spectrum, which is typical for voice services.

Performing the spectral efficiency and traffic engineering calculations results in 1.2 Mbps required to support this service at the cell site – that translates to approximately a single E1/T1 line.

Similarly, a GSM/EDGE application with 3.5 MHz of spectrum results in approximately the capacity of four E1/T1s. A 3G network based on the HSDPA technology with 5 MHz of spectrum requires approximately 13 E1/T1s’ worth of bandwidth.

While transporting 13 E1/T1 lines to a cell tower is doable it is far from convenient, and with 4G networks the urge for more bandwidth at the cell tower is still growing.

The table in figure 4 shows the required services and bandwidth at the cell tower in respect to network evolution.

^{figure 4}

With the increasing adoption of broadband data services, traffic loads in mobile networks are rising enormously.

Mobile Backhaul Transport

The mobile backhaul transport requirements are defined by three primary factors:

• Wireless 2G/3G/3,5G/4G standards
• Performance metrics (latency, jitter, availability)
• Cell site capacity requirements

Current backhaul networks are very well capable to cope with the first two requirements as the radio networks are being prepared to meet mobile broadband demands. Many existing 3G radio access networks have are being- or have been upgraded to meet HSPA capabilities. Further dramatic increases in radio site capacity will come with the implementation of new wireless technologies such as WiMAX and LTE.
Leaning heavily on E1/T1 lines the industry uses TDM/SONET for their backhaul networks as they offer very good performance metrics. Typical limits for the performance metrics are summarized below.

• Frame delay < 10ms
• Frame delay variation 2ms
• Frame error rate 0,0001%
• Service disruption 50ms
• Network availability 99,99%
• Mean time to repair 2hr

As most of the new bandwidth, required at the cell site, can be asychonous as most of it is data instead of speech, investing in expensive TDM networks isn't logical. On the other hand building a new infratructure for asynchonous data offload only is not  desirable either. What if we could have the best of both worlds.
 

Data offloading

While cellular operators are migrating their networks to Ethernet only, they have to choose their equipment very carefully.
Supreme Carrier Ethernet devices are need that at least provide MEF services as:

• EPL, EVPL, EP-LAN, EVP-LAN, E-TREE and EV-TREE
• SLA using enhanced OAM capabilities
• QoS to prioritize traffic
• Resillience: Enhanced protection schemes like ERPS with Y.1731 OAM signalling and end-to-end service protection based on OAM.
• Scalability: very thin granularity of bandwidth allocation.
• Clock synchronization/distribution

Pure ethernet is not the only way of transport in the mobile backhaul. Also IP VPN's or MPLS, MPLS-TP, VPLS can be found in these backhaul networks.
LTE backhaul transport infrastructure can be built using different technologies with layer 2 or layer 3 capabilities. Determining the most effective and efficient mix of layer 2 and layer 3 in the backhaul network is a major issue worldwide. Some use layer 3 routers at the cell sites, and others are strongly opposed to that and want to keep backhaul as simple as possible by using Ethernet. Many operators want to keep as much of their backhaul processing as possible in layer 2, while recognizing that MPLS, MPLS-TP, pseudowires, etc., have elements of layer 2.5. Many operators believe the principal layer 2 advantages over layer 3 are simpler equipment and operations—hence lower equipment and operations cost for a lower cost-per-bit as network scales to support large capacity growth.
Carrier Ethernet with MEF 6 and MEF 8 services can support LTE and LTE-Advanced as well as legacy 2G and 3G, which won’t disappear any time soon. Nearly all operators going to LTE are also moving to carrier class Ethernet as their backhaul transport, because of its inherent capacity improvement and opex reduction advantages.
 

Quality of Service

Capability planning is essential to enable multiple mobile operators to share the same cell tower, and it also enables mobile services with configurable QoS metrics that will be used in future radio access network implementations.
A diagram illustrating the main building blocks of a traffic management implementation in line with the Metro Ethernet Forum (MEF) recommendations is presented in Figure 5. The main functionalities include traffic classification, policing, queuing and scheduling.

In a multi-service environment where QoS differentiation is enabled, mobile operators will clearly want to be able to monitor their packet-based services for faults, continuity and performance. To keep their mobile networks alive and maintain maximum quality, mobile operators will be particularly interested in having the necessary tools to monitor the status of their packet-based services so they can localize faults and trigger corrective actions from remote locations.

^{figure 5: QoS in a mobile backhaul environment}

Ethernet OAM and fault management

The Ethernet OAM standards IEEE 802.1ag and ITU-T Y.1731 provide mechanisms for connection monitoring and performance measurement on an end-to-end service level. Based on the hierarchical concept shown in Figure 6, 802.1ag defines the following OAM tools:

• Connectivity Check,
• Loopback, and
• Link Trace.
   

^{figure 6: 802.1ag and Y.1731 connection monitoring levels}

These tools make use of specific Ethernet frames with unicast and reserved multicast destination addresses. In all cases, these frames follow the same path as the frames belonging to the monitored service. This has the additional advantage that no explicit interworking with Spanning Tree Protocol (STP) or other dynamic network changes is required for compatibility.
Y.1731 builds on 802.1ag to add in performance monitoring features on an end-to-end service basis. For fault management, the following tools are supported:

• Ethernet Alarm Indication Signal, and
• Ethernet Remote Defect Indication.

In addition to enhancements for fault indication and diagnostics, the mechanisms defined in Y.1731 enable the service provider to measure both one-way and round-trip parameters for

• Frame Delay Measurements,
• Frame Delay Variation Measurements, and
• Frame Loss Measurements.

Carrier Ethernet OAM functions for connectivity fault management and performance monitoring provide backhaul network operators with a complete set of tools for assurance and reporting of Service Level Agreements (SLAs) to the mobile operator. While backhaul network operators typically use these mechanisms on a constant basis, mobile operator may choose to only verify the service quality periodically.
A more elaborate description of the ITU-T Y.1731 performance measurement tools can be found in my blog about this topic.
 

Circuit Emulation Dilemma

As the installed base of legacy 2G/3G networks is huge most carriers will not migrate to a packet-only based network. Also, backhaul providers do have multiple customers with different needs. So there will be a demand for legacy E1/T1 connections. One way of transporting both synchroneous and asynchonous data over the same ethernet network is by using Circuit Emulation Service, also called CES.
For many service providers, E1/T1 CES allows them to transition to all-packet networks, while still supporting legacy services. However, CES has its own set of performance issues, which are not acceptable to many wireless service providers.

Circuit emulation involves a trade-off between latency (delay) and bandwidth efficiency. Delay through the network can be reduced, but at the cost of lower efficiency. Likewise, efficiency can be improved, but with longer delays. This concept is shown in figure 7. single T1 frame is transported inside of a single Ethernet frame. The delay is very low, since only a single T1 frame (125 us) is transported.

^{figure 7: Longer delays improve efficiency}

The trade-off is based on how many E1/T1 frames are stuffed inside a single Ethernet frame. If a single E1/T1 frame is transported in an Ethernet frame, the delay is very low. However, the efficiency is not very good due to the CES overhead bytes, Ethernet overhead bytes, preamble bytes, and interframe gap. The alternative is to stuff many E1/T1 frames into a single Ethernet frame. This minimizes the impact of the overhead bytes; however, the latency is much longer due to the fact that 4, 8, or 16 E1/T1 frames’ worth of information must be buffered prior to transmission.

Circuit emulation services are typically 50 percent efficient, due to all of the overhead information transmitted with each Ethernet frame (CES header, Ethernet header, preamble, interframe gap). Many wireless service providers are uncomfortable with the latency, jitter, and efficiency issues related to E1/T1 circuit emulation. For these wireless service providers, their insistence on carrying TDM services in native TDM format is very understandable based on these performance metrics.
Along with CES efficiency and latency also clock synchronisation is an important issue. Circuit Emulation is usually based on adaptive clocking to extract the E1/T1 payload in a synchronized manner from the received Ethernet packets.
The adaptive clocking mechanism however is very sensitive to jitter (delay variation) and many operators might require a more robust clocking mechanism.

Circuit Emulation (CES) allows TDM services to be carried over an Ethernet network, but at the cost of higher inefficiencies and larger network delays. These issues can be minimized by implementing CES when the network traffic is primarily in native Ethernet/IP format, with only a small percentage of traffic in TDM format. The CES inefficiencies and increased delays would then only affect the small number of TDM circuits, and since these would be limited to a small percentage of the total traffic, the overall network impact would thus be negligible. For example, if 90–95% of the network is native Ethernet traffic, performing CES on the remaining 5–10% of services is not a major problem.
As the jitter, or time delay variation, in the mobile network network is very important, clock synchronization is a key issue in the network.
 

Clock synchronization

As said, clock synchronization is extremely important in mobile networks for a successful call signal hand-off between base stations in addition to transport of real time data. GSM and UMTS base stations must support an accuracy od +50 PPB (Parts Per Billion) over 10 year’s equipment life. A drift will result in high dropped call rates and impaired data services.
Currently the mobile backhaul is done via a synchronous TDM network, from which the timing information is recovered.
When the mobile backhaul network is upgraded to Ethernet, the base stations are isolated from the synchronization info that used to be carried over the TDM feeds. A stand-alone solution such as GPS re-timers in each base station is expensive and is rarely perceived as a conceivable solution by most of the mobile operators. As of today, there are two leading methods for timing synchronization over an Ethernet network, ITU-T SyncE (G.8264/G.8264) and Precision Timing Protocol, IEEE 1588v2.

In my next article I will cover both protocols.