The US mobile backhaul revolution

16 October 2012 |

The advent of LTE fourth-generation mobile networks has triggered a revolution in mobile backhaul facilities. Gary Kim investigates the key technologies and players behind this latest trend in the US market.

LTE has been the most consequential development in the US mobile service provider market in 2012. It’s not surprising then that the need to supply LTE mobile backhaul has emerged as a major priority, bringing with it a new focus on areas such as ‘small cells’ and ‘big bandwidth’.

Both developments represent a change in network architecture, an order of magnitude increase in backhaul bandwidth requirements, a shift in capital spending priorities and a historic change in mobile backhaul protocols.

In fact, for the first time that anybody can remember, T-Mobile USA, which in August 2012 upgraded virtually all its backhaul to high-capacity Ethernet connections, is now touting its backhaul prowess in its consumer advertising.

T-Mobile USA boasts a 30Mbps to 50Mbps connection between nearly every tower in T-Mobile’s network and the core network, according to its SVP of technology strategy, Dave Mayo.

“We’re close to being done with our backhaul programme,” Mayo says. That means T-Mobile USA has upgraded more than 30,000 tower sites. “Our aspiration is to have 37,000 modernised sites,” Mayo says, with just “another couple of thousand sites to go, and a handful that we need to find an alternative solution for”.

Perhaps unusually, T-Mobile USA has done so using a variety of ‘alternate’ backhaul providers including cable TV companies, power and light companies. Note the language: Mayo says T-Mobile USA “even” used “some” local exchange carriers. In years past, nearly 100% of backhaul connections would have been sourced from local exchange carriers (telcos).

Other top-four US mobile service providers are also ramping up their modernisation programmes. Verizon Global Wholesale has stated a goal of supplying fibre to 90% of the cell sites within its footprint, approximately 3,500 locations, by 2014.

AT&T has also been aggressively deploying fibre to cell sites, delivering 65% to 75% of AT&T Mobility traffic over fibre facilities by the end of 2011, up from around 20% in late 2010.

By mid-2012, AT&T CEO Randall Stephenson points out that AT&T has “80% of our total mobile data traffic on Ethernet backhaul.” That doesn’t mean 80% of the sites have been upgraded, but more that the sites representing 80% of the traffic have been upgraded to Ethernet facilities.

Sprint Nextel’s vice president of strategic programmes, Marty Nevshemal, says the company’s Network Vision plan includes replacing backhaul connections from leased T1 connections to Ethernet connections.

This forms part of a broader $4 billion to $5 billion overhaul of Sprint’s networks, with the company consolidating separate radio infrastructures for 800MHz services and 1900MHz services and adding LTE. Suppliers for Sprint’s remaining 38,000 cell site connections include virtually all the major US cable operators, AT&T and CenturyLink, while Verizon also supplies a smaller number of connections.

Sprint’s new backhaul is about 90% based on leased optical fibre connections, though microwave is used for about 10% of connections, according to Nomura Research.

Backhaul economics

The LTE-driven backhaul cost implications are substantial. Sprint executives have noted that the company often has to pay $1,500 per month for T1 backhaul at a macrocell tower site, for a total of about 4.5Mbps of backhaul capacity (three T1s). When Sprint switches to Ethernet, Nevshemal says that for the same price of $1,500 per month, Sprint will get almost 20 times the backhaul bandwidth at that location.

The big difference between mobile backhaul for second generation networks and third generation networks so far has largely consisted of an upward bump in the number of T1 circuits typically required to support a single tower site.

Where some cell sites might have operated with a pair of T1s in a 2G setting, a 3G tower might require three to six T1s.

But 4G is a qualitatively different challenge, requiring not 4.5Mbps of total bandwidth, but 50Mbps to 100Mbps or even 150Mbps of backhaul bandwidth. At that level, it is cheaper to buy a single Ethernet circuit than many T1s. In fact, in many markets, it can actually cost less to buy 50Mbps to 100Mbps of capacity than to buy three T1 circuits.

But price is not the only consideration. With the switch to IP bandwidth, Ethernet connections simply operate with more elegance. Mobile operators built out their 2G backhaul networks to support circuit-switched voice services using narrowband TDM backhaul.

But those backhaul mechanisms are not optimised for packet data services. As a result, 3G services launches often forced mobile operators to double the number of leased lines or microwave connections at some cell sites, according to Patrick Donegan, a senior analyst at Heavy Reading.

Small cells used primarily in high traffic areas will pose new issues for backhaul, in large part because the coverage areas are so small that traditional backhaul methods will be too expensive.

A picocell deployment in a sports venue or shopping mall might have similar bandwidth demands. But a public femtocell will often serve a much smaller number of users, with reduced bandwidth requirements sometimes, or often, satisfied by backhaul connections that must cost less than a typical macrocell or big picocell scenario.

Considering that when C Spire Wireless selected Alcatel-Lucent to build and deploy the initial phase of its 4G LTE network, covering a population of 1.2 million and equipping more than 360 cell sites, C Spire invested $60 million for base stations, IP mobile backhaul for 4G LTE and existing 3G CDMA traffic, wireless packet core (WPC) and a complete IP multi-media subsystem (IMS) network core.

That amounts to about $167,000 for each site in equipment costs alone. That is perhaps two orders of magnitude more than a mobile service provider likely can afford to spend on gear to support a single small cell in a high-traffic area. Public small cells might cover areas as small as a typical living room, so that capital investment and backhaul costs have to be scaled back as well.

Where a single C Spire macrocell might be expected to support as many as 3,333 people, a small cell might be expected to handle usage by only scores of people at a single time.

While a macrocell might have a backhaul link of 100 to 300Mbps, each microcell or picocell requires less. “A small cell is going to be somewhere in the 20 to 50Mbps [range],” says Scott Knox, director of Overture Networks Solutions Development.

Small cells to play large role in backhaul

At the end of 2011, there were 283,385 mobile cell sites in operation in the US market. But experts forecast that small cells could quickly equal that figure.

One research firm expects the number of small cells deployed by mobile service providers will overtake the number of macrocells during Q4 2012. That might suggest an additional 284,000 small cells could be added in the US mobile market. In other words, the number of cell sites could double in 2012 alone.

So scale is an issue for small cell backhaul as well. Simply put, small cells might represent an order of magnitude more sites than the macrocell network. On a global basis, Strategy Analytics predicts some 8.6 million small cells will ship in 2017. By 2017, Mobile Experts estimates 66% of all small cell deployments will support LTE deployments.

Nick Marshall, principal analyst at ABI Research, estimates total indoor and outdoor small cell shipment numbers reaching 4.55 million by the end of 2013. It is perhaps not a surprise then that research firm iGR estimates demand for mobile backhaul in the US market will increase 9.7 times between 2011 and 2016.

The transport piece of backhaul is expected to grow at a 2% compounded annual growth rate to $6 billion by 2016. Microwave will be used to backhaul over half of the sites, according to Dell’Oro Group’s VP, Jimmy Yu.

Implications for mobile backhaul suppliers

The largest mobile service suppliers source their backhaul services in a number of ways. Both AT&T Mobility and Verizon Wireless can rely on landline assets owned by their respective parent companies, inside their respective service footprints.

Neither firm has such assets in more than a fraction of the US though, so AT&T and Verizon Wireless are major buyers of mobile backhaul services outside the areas where they operate fixed local networks.

Sprint, T-Mobile USA and the Tier 2 mobile service providers operate no local networks, so must buy 100% of their mobile backhaul capacity.

Given the importance of backhaul costs, suppliers other than the larger traditional telcos have for the first time emerged as an important source of supply, though AT&T, Verizon and CenturyLink remain key suppliers of backhaul capacity.

CenturyLink’s backhaul operations are handled by its Wholesale Markets Group, which completed approximately 1,350 fibre builds during the second quarter of 2012 and over 2,000 links for the first half of 2012. The group ended the quarter with about 12,150 fibre-connected towers and remains on track to complete 4,000 to 5,000 fibre builds in 2012.

The leading US cable operators in 2011 served or had under contract at least 18,000 cell towers, or about 7% of the nation’s 253,000 cell sites, according to Heavy Reading.

Comcast and Cox Communications are among the US cable operators which have seen a big upside in mobile backhaul. In late 2011, Comcast already had contracts serving about 5,000 towers, while Cox had sold services serving about 3,500 towers.

And Comcast notes that the market opportunity for AT&T Mobility and Verizon Wireless is about 100,000 mobile towers operating in its service territory.

Comcast sees mobile backhaul as potentially a $1 billion business for the company over time. Mobile backhaul was a large chunk of Cox Business’ approximately $100 million in 2010 wholesale transport revenue.

Time Warner Cable, which saw growth in mobile backhaul revenue increase from $7 million in 2008 to more than $60 million in the first half of 2011, is also bullish on the business.

For Cox Business, wholesale transport accounted for $100 million in 2011 revenue, with cell backhaul representing the highest growth area. Time Warner Cable’s Q4 2011 cell backhaul revenue was $26 million last year, which was equal to its total amount of cell backhaul revenue in 2009.

In 2011, cable operators already passed most existing cell towers with either hybrid fibre/coax (HFC) or direct fibre lines. Indeed, some analysts believe 80% to 90% of all existing US cell towers are located within cable’s existing network footprint. Globally, mobile backhaul market revenues are projected to top $8 billion by 2015, according to a recent report by the Dell’Oro Group.

Level 3 Communications, Zayo Communications, Telecom Transport Managment (TTM), FibreTower and Tower Cloud are some of the independent companies that also compete with Verizon and AT&T to provide mobile backhaul services to mobile service providers such as Sprint and T-Mobile USA.

FairPoint Communications, Frontier Communications and Windstream are among the Tier 2 local telcos now selling mobile backhaul connections in their service territories.

Almost any way one looks at the matter, LTE has triggered a wide-ranging evolution of mobile backhaul in the US market, displacing older protocols, bandwidth assumptions and cell architectures. Small cells and a switch to Ethernet connections of 50Mbps to 150Mbps are prime examples.


Technical talk: Knowing your picocell
from your femtocell

Some idea of the magnitude of the changes to the backhaul business can be gleaned by considering just the types of connections, cell architectures and bandwidth requirements. Fourth generation networks such as LTE use Ethernet connections, with links routinely provisioned at 50Mbps to 150Mbps.

According to Mike Schabel, VP of lightRadio programmes at Alcatel-Lucent, network requirements have evolved during the transition from 2G to 4G: “The 4G network is purposely designed for jitter and latency performance to support gaming and video. The earlier networks were voice driven,” he says. “The mindset has changed. The business used to be voice driven, with a little data, and YouTube was a novelty.”

Now, however, video is a mainstay application to be supported. New 4G networks are therefore expected to make heavy use of small cells, a development with key implications for backhaul. Even defining small cell, however, can prove daunting.

For Schabel, small cell encompasses every cell deployment “smaller than a macrocell”, including both indoor and outdoor solutions, business or residential, public network or end-user deployed. But “smaller than a macrocell” includes backhaul scenarios of varying bandwidth requirement.

As a general rule of thumb, small cell backhaul has to cost less than macrocell backhaul. Also, small cell bandwidth demand will be ‘burstier’ than a macrocell, given the heavier expected internet access demand at a small cell site, where users are ‘on foot’ rather than in a moving vehicle, and therefore more likely to be using internet apps.

Small cell performance requirements, for example, also tend to be less stringent than for a macrocell in the latency and quality of service (availability) areas. A microcell often covers a limited area such as a mall, a hotel or a transportation hub. Typically the range of a microcell is less than two kilometres wide, while a picocell is 200 metres or less, and a femtocell is on the order of 10 metres.

The backhaul requirements for a microcell might not be substantially different from that of a macrocell. A picocell, on the other hand, more commonly serves a high-traffic area that is relatively small, and might reasonably be supported by backhaul connections featuring bandwidth more normally associated with a digital subscriber line connection (less than 45Mbps but perhaps more than 10Mbps).

A public femtocell might serve so few potential users that nothing more than a DSL connection makes financial sense. Small cells therefore span a range of deployments for either coverage or capacity, with direct implications for the amount of backhaul bandwidth required, as well as the permissible cost of such connections.

Coverage scenarios include indoor venues or areas of signal shadow, says Peter Linder, Ericsson director. Capacity scenarios more often will include usage hot spots in dense urban areas. Still, from a backhaul perspective, Linder says it is “hard to draw hard lines between femtocells and picocells”.

Generally speaking, however, macrocells have the highest requirements for guaranteed quality of service, while many small cells will provide value even with “best effort” QoS, according to Ceragon Networks, a supplier of mobile backhaul systems. Macrocells are always designed for “call or session handoff.”

In many cases, small cell sessions will not have to support session handoff, as sessions will start and end within a single cell’s coverage area. Also, in terms of spectrum and bandwidth management, macrocells will support multiple frequency bands and multiple “sectors.” Small cells will often only require support for a single band. And while macrocells necessarily are optimised to support voice applications, small cells often will feature a primary emphasis on data sessions.