Mobile networks that optimise themselves could save operators money, E&T investigates.
Mobile operators are in a bind. The uptake of smartphones has led to an unprecedented increase in the volume of data flowing over their networks. They've been scrambling to keep up, investing in capacity even as they consolidate their 3G networks and make ready for a shift to 4G technologies.
Speaking at the Mobile World Congress in Barcelona earlier this year John Donovan, chief technology officer of US network operator AT&T, put the problem bluntly: 'More apps means more bits. The growth in usage of our wireless network in the fourth quarter of 2009 compared to the third quarter of 2009 was greater than all the traffic we carried in 2007.'
His response to this runaway growth has been to push hard for efficiency gains: he says the cost of carrying one extra bit of data on the AT&T wireless network halved last year.
'We have to wake up every morning and sweat efficiency,' he said. 'We have to stay on that curve and pursue it relentlessly.'
Building more basestations is one way of improving coverage, but the industry is also on the verge of moving to LTE, the most broadly accepted 4G technology, in a bid to squeeze more bandwidth out of the limited spectrum available for mobile communications.
Frédéric Pujol, mobile broadband practice manager at market analysts IDATE, has forecast that there will be 380 million LTE subscribers by 2015, with operators such as Verizon and Telia already trialling the technology in cities.
The emergence of LTE is a double-edged sword for the operators. A more spectrally efficient radio technology will help them meet rapidly rising demand for bandwidth from limited spectrum resources. On the other hand, the rise of LTE means that operators face another round of investment to service a brutally competitive market in which consumers have been taught, thanks to flat-rate deals, that they can consume as much data as they like for a fixed sum. And then there's the complexity overhead.
Ajay Khanna, chief technology officer of network optimisation software company Celcite, said: 'Eight to ten years ago, operators had one network. Six years ago UMTS came in, followed by more carrier frequencies, and operators will soon be introducing LTE. They are now using multiple infrastructure vendors. Maintaining multi-vendor, multi-standard networks is a nightmare.'
If the operators' investment burden is rising and yet their ability to recoup that investment is under threat from competition and consumer expectations, something has to give, in this case the costs of buying and running their networks. This has led to rising interest in self-organising networks (SON), which can configure themselves, optimise themselves in operation, and heal themselves.
The Next Generation Mobile Networks (NGMN) Alliance, which exists to ease the transition from current to 4G networks, is already concerned about the operating costs of the new networks, and so is pushing equipment vendors to come up with systems that can manage themselves, as well as interoperability at the network and service-management level so that equipment from different vendors can collaborate to optimise overall performance.
As a first step, NGMN is calling for plug-and-play basestation set-up and parameterisation, as well as automatic neighbourhood planning for adjacent basestations. To aid progress towards more automated networks, the Alliance has a list of 'use cases', such as cell-to-cell handover and automatic set-up, in which it believes automation can play a part.
'Self optimisation needs vendor interoperability and the NGMN Alliance can act as a forum for this,' a spokesman said in Barcelona.
Now that the use cases are defined, each can be addressed as a separate issue. Some of them, such as neighbourhood optimisation, have already been addressed in UMTS networks. For example, the Vodafone Sure Signal home basestation (a 3G femtocell) will configure itself once it has been given its location and height above ground.
'It's about codifying, standardising and broadening the answers to these known problems,' says Keith Reed, director of network analysis products at network optimisation company Actix.
He argues that a lot of self organisation is about creating closed feedback loops. The question then becomes, at what level of the network hierarchy should these loops be made? Some loops can be closed in the hardware, so that it adjusts itself based on information that it gets from mobile users and neighbouring cells, a form of distributed self-optimisation that infrastructure makers can develop. But more complex optimisations, such as balancing loads between cells, may need to be managed in a more centralised way using a larger closed loop.
Such schemes will rely on the equipment buyers getting the equipment vendors to follow standards for various forms of network information such as current configurations, current performance and diagnostic messages, so it can be used by third-party management applications. The uptake of SON will also take a change in attitude from the operators themselves, who may be reluctant to accept too much automation too soon.
'If you show operators closed-loop solutions now they won't buy, at least not for every use case,' says Reed. In a case such as traffic overload management, a software tool could take the overload data, highlight the problems it was causing and offer solutions that could be manually implemented.
'It is not a closed loop yet,' says Reed 'though after a number of optimisations, the tool could offer to do [the optimisation] automatically.'
For the tool vendors, of course, the more faith the operators put in their tool and the more deeply they embed their network management policies within it, the better.
'Mobile networks need to get more Belkin and less Ericsson.'
So how can networks optimise their performance? According to Christopher Millhouse, business development director at cell-planning software company Symena, some of the most powerful optimisations involve physically moving the basestation antennas to alter their propagation pattern, or footprint, on the ground.
Modern basestations can offer remote electrical control of the tilt and direction of their antennas: 'This is the most powerful way of shifting things around the network.'
This approach can be used to balance the load on a basestation between over- and under-loaded cells by changing the position of the overloaded antenna to shrink its footprint and therefore the number of handsets it is trying to serve. Traffic from handsets that can no longer reach the overloaded cell will roam to less loaded cells, automatically balancing the load. It's also possible to adjust the power on each antenna but this is not considered as effective.
'A large percentage of the change possible can be picked up with remote electrical tilt,' says Millhouse.
He sees increasing opportunities for his company's products, which try to choose the best position and type of antenna to use on each basestation to serve the predicted traffic, because traffic patterns are changing so rapidly. For example, he says that one Japanese CDMA operator changes its network once a week. In the UK one operator misforecast its traffic patterns, expecting that traffic would be strongest in cities between 11am and 12pm and later finding that it was actually peaking after 6pm, when people got home from work and started using mobile broadband.
Macro versus femtocell
Self-organisation will also be important in LTE networks because of the likelihood that many networks will be built of a combination of macrocells to cover large areas, and pico and femtocells to fill gaps and service hotspots. Femtocells, which have been designed down to consumer prices so they can be sold as a way of improving 3G coverage in homes (like the Vodafone Sure Signal box) are now being re-engineered to handle tens of simultaneous users in public LTE networks.
Simon Saunders, chairman of the Femto Forum, sees a 'particular opportunity for rapid deployment of femtocells in the outdoor arena to soak up capacity hotspots', arguing that femtocells can help provide high-quality LTE connections early in the roll-out of 4G. Others argue that femtocells will help extend LTE to rural areas and other places where the likely traffic wouldn't justify the cost of a macrocell.
But femtocells, especially for use in homes, have to be able to set themselves up as well as taking their place in a wider network of macrocells. As the NGMN said: 'If you have to have an engineer setting this up it's not going to fly'.
Will Franks, founder and CTO of Ubiquisys, which provides femtocell reference designs, suggests that the problem is already being tackled.
'What we did at 3G maps to LTE principles,' he says. 'The reality is that in any size of enterprise building, femtocells have to be installed by IT-qualified people not RF engineers. They need to be able to self-organise and self-optimise. So we looked at Wi-Fi planning, where you put a coffee cup on a map of the building [to work out how the basestations should be placed]. Based on that approach, you don't need to send your RF engineers in to take up the ceiling to put RF cables in.'
Of course, the introduction of a new network technology offers plenty of opportunities for profit, plenty of opportunities to shake up the established order - and plenty of opportunities for conflicts between the interests of suppliers and customers.
Keith Reed at Actix puts it this way: 'One of the things about new technology is that it's an opportunity for musical chairs in the equipment makers. The business model of Asian equipment vendors is not so service-driven as it is in the US and Europe.'
Why is that relevant here? The argument goes that some equipment vendors have seen the writing on the wall in terms of the commoditisation of their hardware offerings, and so have been developing higher margin services. If the hardware starts being able to set itself up, optimise its operation, coordinate with the rest of the network and heal itself, lucrative service revenues could be under threat.
Khanna points out that companies such as Ericsson and Nokia Siemens Networks make more than 50 per cent of their revenues from services, and he expects infrastructure equipment revenue to shrink to 20 per cent of revenues. Asking companies whose revenue is structured this way to increase the automation of their equipment means 'asking equipment makers to do something not in their interest'.
Millhouse at Symena makes a similar point: operators often specify basestations to have enough capacity to support the traffic in their busiest hour.
'If you have a self-optimising network you don't have to build to the busy hour,' he says, because local optimisations will help an overloaded basestation out. That's not going to be music to the equipment makers' ears.
He points out that conflicts aren't limited to relationships between customer and supplier. For example, Ericsson is now among the top 10 operators by customer numbers by virtue of its role as a network manager, which creates a potential tension between the service and the manufacturing business.
The NGMN Alliance is pressing the equipment makers to provide the hooks necessary in their hardware to enable self-organisation, optimisation and the rest. But some are sceptical that the different vendors will comply rigorously enough to any standards set to make it easy for basestations to collaborate.
'Our opinion is that self-optimising networks will only be big in LTE, because the NGMN Alliance requirements give you a whole lot more hooks for data,' says Millhouse.
Potential difficulties about incompatible implementations of standards will create an opportunity for software vendors to position their tools as providing a centralised way of ironing out incompatibilities, handling different interpretations and imposing optimisations that equipment vendors are reluctant to implement themselves. Millhouse describes it as conducting 'a self-optimising network orchestra'.
'The thing about automation is that it takes a lot of people to make it happen,' he says. 'A lot of things have to come together, which is why we see it as an orchestral effort.'
Khanna says his product, (COPS) can act as an abstraction layer that overcomes the vendor differences and offers a 'dashboard' that presents all the data, does intelligent correlation and provides an analysis of what needs to be adjusted.
'We're trying to fill the gap that self-optimising networks will create,' he added. 'Even if [operators] have self-optimisation on their vendor equipment, we know for sure it won't go across UMTS and LTE.'
There's a huge opportunity to use automation to cut the cost of delivering current and next-generation mobile networks, although there's a tension in the provision of the equipment and the evolution of the enabling standards between those companies that just want to sell kit and those who want to sell services as well. In the longer term, though, this may become a moot point. Advanced LTE networks are likely to use techniques such as beam forming, coordinated multipoint antenna schemes and adaptive transmission schemes to squeeze even more usable bandwidth out of the spectrum. Making such schemes work will take such close coordination between basestations that self-optimisation will become a necessity, not an option.