Analytics/Big Data

Big Data Lights Up the SON

During my 25 years working in the wireless industry, there has never been a shortage of technology change but my experience over the last couple of years has demonstrated a marked increase in the complexity of deploying and operating wireless networks.

The exponential growth in data traffic with resulting capacity challenges is forcing Operators to use all tools at their disposal to manage the traffic increase. Network architectures, by necessity, are becoming much more layered and complicated and this will likely continue for the foreseeable future.

The wireless network environment is undergoing a revolution with planned massive deployments of small cells that are both driven and enabled by LTE. Operators that have managed a few thousand cell sites must now create processes to manage tens of thousands of sites, all of which will constantly be changing. As two examples, Verizon Wireless has recently reported capacity issues and stated small cells as a specific element of their solution, and AT&T Inc. (NYSE: T) has stated that they plan on deploying 40,000 small cells in the future.

Furthermore, the expanding array of wireless devices and applications accessing the networks is also challenging operations management. Operator core networks are changing to support the influx of different services by implementing new platforms that are far beyond traditional wireless equipment. I have already seen Network Operations organizations working full-out just to keep up with current demands, and it is clear that the coming exponential growth will not be able to be handled by simply throwing more bodies at the issue. The net result is that without strategic technology enhancements, networks will very soon start outpacing the operators' capabilities to manually manage and optimize deployed assets, and correspondingly, the end-user experience.

I believe there are two technology solutions now coming to the forefront and, when combined, these solutions will become powerful tools necessary for operators to meet the complexity challenge. The first technology is the Self-Organizing Network (SON) and the second is big data analytics. Overall, the combination of these two developments permit operators to introduce automation allowing efficient deployment and management of significantly larger number of network elements.

Wireless operators are now beginning to deploy the first versions of SON technology. SON holds the promise of creating a set of functionalities that allows the automation of many operations processes, reducing the manual effort required and increasing the quality of those processes. One SON application being deployed today, called Automatic Neighbor Relation (ANR), enables a cell site to self-configure its acceptable neighbor list based on information it receives with minimal to no human intervention. Therefore, a cell site installation process that previously could have taken hours to perform manually can now happen, autonomously, in minutes.

AT&T has been an early proponent of SON and has publicly reported major operational improvements with its initial SON use. (See The SON Always Shines on LTE and AT&T's No. 1 SON.)

The second technology that will need to be added to the mix is the appropriate and timely use of data analytics. In the classic sense of "garbage in-garbage out," the benefits of SON automation will only be maximized to the degree that the information used to trigger and implement a response is highly accurate. Achieving improved input data will often come about by correlating information from a wider set of sources that, individually, provide different perspectives on a particular situation.

The interest and expectations for SON and big data are already very high. However, as we have seen throughout history, the implementation of capabilities that radically alter the operational environment tend to be more evolutionary than revolutionary. Therefore, the expectation is that the introduction of SON will happen pragmatically and incrementally as experience and trust levels grow. Two primary dimensions will likely drive the actual evolution of SON capability. The first dimension defines three types of operator activity impacted by SON: (1) self-configuration, which often occurs during the initial implementation of a new element; (2) self-optimization, where the element modifies its operation based on new information, learning or long-term network changes; and (3) self-healing, where the network elements react to particular disruptions, outages or environmental situations and minimize impact to end users.

The second dimension centers on the type and consistency of network elements involved in a SON process. In operations today, SON approaches usually address a single network platform, such as the eNodeB, within a single vendor's domain.

In time, SON solutions will likely involve more standardized interfaces and accommodate multi-vendor solutions. The migration will also move beyond operational platforms to involve business support systems, resulting in a holistic integrated ecosystem. In some configurations the variety of data sources will allow new use cases from different inputs. Often, existing use cases will gain incremental improvements from more refined and appropriate decisions based on deeper knowledge sets. Overall, operators benefit by optimizing their entire networks rather than individual network segments.

I think that operators are still very nascent in implementing the combination of SON and big data, however, experience gained in the near future will allow for more full scale introductions in the future. My sense is that to maximize SON optimization efforts, the data analytics services in these situations must be able to process billions or more transactions from a wide variety of data sources and provide actionable outputs in near real-time.

Achieving this is possible but not simple, requiring a very concerted and cross-functional effort that may be disruptive for many operators. However, those operators who are successful in implementing SON and Big Data will be able to manage their growth most efficiently and also, potentially, provide a point of differentiation in an increasingly competitive world.

— Rob Chimsky, VP of Insights, Guavus Inc.

rchimsky 12/16/2013 | 3:30:40 PM
Re: Related costs Got it and absolutely agree.  The challenge right now is that the clearest monetization ideas tend to involve sharing analytics with third parties such as well documented opportunities around mobile advertising and targeting.  Unfortunately, the heightened sensitivity around privacy will probably tend to limit use cases in this fashion for some period of time.  However, there are the potential use cases for operators to use internally, which is where we will likely first see major efforts.   Clearly there are uses for improving overall customer experience and importantly helping to manage churn...although every carrier business case tends to involve managing churn and is very hard to prove so Operators will look for more concrete justification.   The types of services they can look at internally may be real time QoS type policies where a user can pay incrementally to assure better experience for specific applications.  Everyone is still learning what is possible here and what will ultimately be successful, so we will probably see much experimentation and more failures than successes in the short term, but as you say, finding the right path to both optimization and monetization are critical for future Operator survival.
MordyK 12/16/2013 | 3:06:55 PM
Re: So also add in a value component You need to account for all the new cells deployed in the form of small cells which will be far more than the number of existent macro cells, as well as teh costs of maintenance.
MordyK 12/16/2013 | 3:04:40 PM
Re: Related costs Rob, I totally agree with you on the value of this. But I'm saying that carriers also need to look at additional opportunities to monetize big data so that they can expand the resources allocated to this and gain value from both SON improvements as well as other alternative sources.
rchimsky 12/16/2013 | 2:48:44 PM
Re: Related costs There is no doubt that Operators will continue to be cost challenged as they look to optimize capex and opex in a revenue constricted environment.  I believe that based on this we will not see what I would call "big bang" moves in these directions, where we see revolutionary changes to the networks or operations.   We will see much more pragmatic moves that involve both SON and Big Data.   Low hanging fruit are things like neighbor list management where the operations savings and quality improvements are significant.  As Operators get comfort from these targeted solutions they will then look to expand.  This will place emphasis on very scalable platforms and implementations.   Also to the question on overall value.  AT&T has probably been the most public about their experience with SON.  In AT&T's most recent annual report they talked about SON and stated that they have seen a 10% reduction in dropped calls because of the efficiencies from SON.  This does not included the manpower savings which have also been signficant.   
MordyK 12/16/2013 | 12:01:31 PM
Re: Related costs Storage, processing, data scientists, etc.
brookseven 12/16/2013 | 11:47:49 AM
So also add in a value component "One SON application being deployed today, called Automatic Neighbor Relation (ANR), enables a cell site to self-configure its acceptable neighbor list based on information it receives with minimal to no human intervention. Therefore, a cell site installation process that previously could have taken hours to perform manually can now happen, autonomously, in minutes."

So how many hours in total does that save a year.  Let's say for grins that 8 hours of work are reduced to 1 - just trying to round numbers here.  This means that on an annual basis one saves 7 x the number of cell site commissioned.

So, unless we are talking about something for 5G then is it of great value?  What percentage of the 4G network is rolled out in the high labor cost markets?


Sarah Thomas 12/16/2013 | 9:58:35 AM
Re: Related costs What kind of costs are we talking about here? Maybe Rob can give us an idea, too. It seems like an investment operators have to make, but they're exploring different ways to implement it -- and store it -- that might reduce costs.
MordyK 12/16/2013 | 9:50:08 AM
Related costs The marriage of SON & big data is obviously a match made in heaven, but the heavy costs of these new capabilities need to be offset without further burdening the subscribers.
Sign In