I'd not been to a SON (Self Organising Networks) conference before and was keen to capture a snapshot of the state of this part of the industry. Would this be seen as a pre-requisite for small cells, or is the focus of this technology elsewhere?
SON software promises to improve the efficiency and capacity of complex HetNet mobile networks, and is increasingly important to handle the rapidly growing number of cellsites and radio technologies/frequencies used. It already demonstrates impressive return on investment, but will be a critical component for large scale small cell deployments.
What is SON?
SON automates many of the technical processes to configure and optimise the Radio Access Network, and is essential to cope with the rapidly increasing complexity – from additional radio access technologies (LTE), frequency bands (3G and 900MHz, LTE at 800MHz, 1800MHz and 2600Mhz), and multiple layers (macro, micro, smallcells).
This year's conference had a different flavour from last year
Attending the conference, we found several 3rd party C-SON (Centralised SON) vendors actively exhibiting and presenting (Celcite, Celwize, Actix, Reverb etc.) There weren't any D-SON (small cell vendors) presenting their side of the story, but the Small Cell Forum presented and at least one RAN vendor (Alcatel Lucent) was there to speak up about their own C-SON and D-SON capabilities.
This year's conference was said to be about the same size as last year (I estimated around 80-100 attendees), but delegates told me they thought it had a lot more structure and a clearer agenda than before. The big RAN infrastructure vendors were mostly absent but it did include speakers from Sprint and Verizon Wireless, Bell Canada and Telefonica.
SON is primarily cleaning up the 3G mess
Stephane Tefal from Infonetics opened the event with a frank assessment that while the industry was busy rolling out exciting new LTE technology, it was in "cleaning up the 3G mess" that you could find most SON activity today. Revenues for SON are growing all the time, with hundreds of millions of dollars being spent annually – mostly on 3G today despite the LTE headlines.
His view is that spending on macrocell equipment will peak this year (2013) with the initial investment in LTE rollouts having brought forward some of the 2014 budgets. Thereafter, there will be a strong shift into ICT (software systems including SON to manage networks more efficiently). Equipment budgets will evolve, with capacity and coverage for indoor and outdoor requiring small cells and picocells.
Case studies proved the value
There were several case studies demonstrating the benefits – one presenter saying "only a lazy guy (radio optimiser) would not be using ANR (Automatic Neighbour Relations) today".
A case study from Cellcom Israel highlighted the extreme pressure that some operators are under. Their regulator had mandated that prices were cut by 50%, leading to 50% staff reductions and equipment budget cuts of a similar magnitude. An Austrian delegate commented that similar regulatory pressures in his country too. Cellcom are said to have only about 20 radio optimisers for a network of around 30,000 cells. (My guess is some networks would have anything up to about 10x that number of people, and such a low staffing level must have some side-effects for efficiency). There simply isn't any feasible way of handling this workload without lots of automation.
A specific graph shown from that case study was the call drop rate over time, which consistently improved using ANR. When senior executives questioned whether the SON system was really required (surely it would have completed the optimisation after a few months), it was switched off for a trial period. After 4 weeks, the call drop rate doubled, but recovered back to former levels within a week of being switched back on again.
Others have told me that C-SON (rather than D-SON) is needed to manage the neighbour lists because the system needs to be able to delete neighbours as well as add them, something only a centralised system would have enough information to be able to handle.
Main business benefits of SON
Key themes emerged to justify these complex software systems:
• CAPEX savings, specifically deferring purchase of additional RAN equipment by squeezing more capacity and coverage out of the existing installed equipment.
• OPEX savings, through fewer staff (despite what some operators and vendors said, I didn't quite believe that there wouldn't be cutbacks where SON systems are deployed)
• Customer Experience improvement, leading to much lower churn rates. A case study from Celwize highlighted that where churn was caused by poor network performance, their approach halved that churn rate and more than justified the optimisation project.
• Scalability, to handle deployment of lots of new small cells, new frequencies and expanded equipment such as sector splitting.
A wide range of SON techniques were discussed, with the most popular being ANR and CCO (Coverage and Capacity Optimisation). These are provided by 3rd party Centralised SON solutions. Chris Larmour, Celcite, confided that the lengthy trial periods previously required to validate their system were rapidly reducing. It seems that CTOs accept there is less risk from these mature software solutions, with one network prepared to go live nationwide only four weeks after an initial trial. He estimates some 500K cells are under Centralised SON control today, and expects this to grow to over 1 million by early 2014.
These issues came up during our analyst breakfast briefing session, which included one table led by ThinkSmallCell.
The periodic table of SON features
Neil Coleman of Actix (now acquired by Amdocs) listed a whole host of different SON techniques and features which have been developed. He illustrated this in a chart format that looked very much like a chemical periodic table (shown below), with 3-4 letter acronyms for each one. It resembled Qualcomm's long list of small cell D-SON features which I'd seen earlier in the week.
Asked how conflicts are handled where different features want to adjust the same parameter in different ways, Neil clarified that any mature SON system would have that wider orchestration capability and oversight to prioritise and resolve these situations.
Where are the bottlenecks?
I asked several speakers what was stopping C-SON systems working faster than the 15 to 30 minute cycle time today. It seems that it's not the huge volume of data being uploaded from the RAN, but the latency of how quickly it can be accessed. Provisioning new parameters down into the RAN can also be a limiting factor because it overloads the Network Management Systems, but once the system has been live for some time, the quantity of adjustments slows to a steady flow than bulk updates.
Some RAN equipment, especially newer models, can provide live feeds rather than bulk data files, which improve the situation.
Regardless of how quickly these C-SON systems operate, there will still be a need for D-SON within the small cells, which rapidly responds to environmental changes.
Is small cell provisioning and initial setup part of the core SON functionality?
Configuration Management, which determines the initial parameters when a new cellsite (large or small) is commissioned, is quite a complex subject. This includes settings which have little to do with radio optimisation, such as determining where to route emergency calls, which network gateway to connect to and which backhaul route to use. Rob Soni, Alcatel-Lucent's CTO for small cells, commented that he thought the initial setup configuration would continue to be handled by each (small cell) vendor's systems. Once online, the SON systems could then integrate and optimise the new cell into the rest of the radio network.
Doug Pulley, representing the Small Cell Forum, illustrated a series of cases where small cells affected macrocell and system performance. These were addressed mostly by D-SON and techniques built into the small cells themselves.
I also raised this issue in the panel session which I participated in because I felt it had been overlooked.
More radical ideas
TMO (a Dutch Research Organisation) discussed a project which is looking at dynamic spectrum refarming. A cellsite might change which radio technology is used at a particular frequency band at different times during the day, sometimes operating GSM and at other times operating LTE. It's the opposite of traffic steering, where handsets are switched to use the best/most appropriate radio interface available. Frankly, I'm a little sceptical of the value this brings, but will try to remain open minded.
Various organisations working on recommendations, guidelines and standards
The Small Cell Forum release program is due to publish major documentation for Enterprise and Public Access small cells in December 2013 and February 2014. In addition, Giovanni Romani of Telecom Italia and Vice Chair 3GPP RAN outlined 3GPP's current Release 12 features which include several Small Cell related aspects. He also covered an upcoming report from the NGMN which will cover small cell requirements and is due out in March 2014.
The key benefits of SON include:
- CAPEX deferral
- Customer experience improvement
- OPEX savings from fewer staff and energy savings
- Provisioning/deployment scalability and speed
I thought this conference mostly focused on the first two points.
SON remains something of an overlooked niche today, perhaps because so many operators have focussed on the headline activity of rolling out their new LTE networks. It can quickly deliver substantial benefits in OPEX, CAPEX and Customer Experience, and is slowly becoming much more accepted especially for large 3G networks. The risk averse and conservative pace of software/systems management technology adoption within operators is being forced by stronger financial pressures.
I heard several software vendors say they'd like to be the next Intucell (bought by Cisco for over $400 million), which has raised the profile of the SON industry significantly. There's also lots of SON activity quietly going on within the major RAN vendors, but 3rd party C-SON systems would surely provide the better multi-vendor oversight in the long term.
I felt there wasn't quite enough concern about the issues of scaling up to handle large numbers of small cells at this conference – it was more related to current macro network topologies and especially 3G. The increasing pressures on operators (especially price regulation in Israel and Austria), have driven greater adoption of SON and I would expect this to grow even more quickly over the coming months.
One operator delegate observed that his senior management refuses to plan for much more than one year ahead, because they feel that nobody can accurately predict what will happen further out. I believe we can take an educated guess on some aspects, but the scenarios aren't straightforward or simple. SON should surely be a part of every operator's future roadmap.
And finally... while I can recommend Nice to be a very attractive place for a conference, sadly this was as close as I got to the beach... the view from the venue hotel balcony.