More on SON

As Mobile World Congress gets underway this week, I expect a lot of noise will be made about Self-Organised Networks along with grand claims and promises how operators can utilise SON to substantially reduce their OpEx. So why has it taken so long? Last time I finished by asking how hard can it be? Let me try and summarise this here.

Ideally to fully realise SON, the networks elements would be 'plug and play'. Apart from the physical act of installing the equipment once it is turned on the network element figures the rest out itself. Additionally, it would be able adapt to changes in the network due to faults or configuration changes to ensure that it continued to function optimally. Networks capable of this would substantially reduce the cost of planning, provisioning, operations, optimisation and upgrade.

Why is this so hard to achieve? If you take a radio access network for example. The air interface is a hostile a environment, interference and physical obstructions are just two hurdles that have to be overcome. SON capable network equipment would need to be able to determine what frequencies are available for them, determine the optimal power levels that can maximise coverage yet minimise interference will adjacent cells and automatically determine its neighbours for handover in order to allow end user mobility. That's just for starters.

Any configuration changes is likely to be subject to a potentially complex set of configuration rules that are dependent on the specific software version running on the network elements. Its unrealistic to expect all network element software upgrades to occur at the same time - never mind the associated risk. Therefore the network will be required to work with multiple software versions running simultaneously, and any proposed configuration change would be to be valid for that specific version and the current configuation of the live network. Given that a complex NE like a UMTS Radio Network Controller (RNC) could have 1,000s of configuration rules, it can all get very complicated very quickly.

The real killer is the network operators expectation for 5NINES (99.999%) availability - approximately 5 minutes downtime annually. Network outages = zero revenue for the duration of the outage and therefore can quickly become very expensive for the operator. If it was easy to automate, it would have been done already.

Therefore the need for meticulously planning and controlled deployment of network changes by skilled personnel which is why the operators are trying to find ways to reduce the cost of Operations Expenditure. So high in fact, that some operators have outsourced the running of their network to managed service providers instead of running it themselves. Despite falling network infrastructure sales, equipment vendors such as Nokia-Siemens Networks, Ericsson and Alcatel-Lucent have all seen growth in their managed service business. In fact NSN have predicted this market globally to grow to €200Billion (See NSN Sees Managed Services as $277B Market) and this only represents the cost where operators have decided to outsource the business to a 3rd party. This gives an insight into the staggering cost of running networks for the operators. Hence, the need for Self-Organised Networks!

Webinars
Attended two webinars last week. The first was Subscriber Data Management: Next Steps sponsered by Blue Slice Networks. Next Generation SDM solution for managing subscriber data. It highlighted the migration away from vertical subscriber data such as in the network (e.g. HLR) or in the back-office (e.g. CRM) towards an evolved SDM. The 3GPP are defining User Data Convergence through specification of standardised interfaces.

The second was Assuring The User Experience for IP-based Mobile Networks from Netscout. Described how a unified approach to service delivery management can be achieved or put another way how an end-to-end IP network can be managed. This will be critical for the operators who are trying to find new revenue streams through the provisioning of premium services. Netsocut solutions use deep packet analysis that can analyse every packet to identify sources of problems. This is very relevant to my current research in EFIPSANS and I think would be worth looking into the capabilities of these solutions more.

In the news this week...
Alcatel-Lucent finally reported its fourth quarter earnings and it was no surprise to see that it also saw a decline in revenues, a trend seen by most but not all of its main equipment vendor rivals. (See AlcaLu Revenues Shrink 11% in 2009)