Wireless network complexity
Wireless networks are becoming more complex. This raises a question: Are we nearing a point where network complexity is moving beyond the capability of humans to manage? Are we already there?
Let’s look at just two data points to get a feeling for the challenge at hand.
First, architectural complexity. From 2G to 4G, we could say growth has been linear and consistent in terms of the number of base station sites and the fundamental architecture of networks. Macro site expansion has followed spectrum acquisitions and sites have been reused to host multiple technologies. While certainly challenging, growth seems to have been managed well by operators, as manifested by consistent and improving end-user services quality.
However, going forward, we envision networks becoming inherently more dense and complex in makeup. Part of this will come from 4G, where some operators are planning to add more small cells than the total number of BTS macro sites they already have. The other part will come from “5G,” where cell counts are likely to increase by perhaps an order of magnitude as higher frequencies are used, changing cell sizes from several kilometers to a few hundred meters at best. Adding to the numbers is the makeup of these additions. In both cases, small cells are more likely to be placed on poles rather than rooftops and towers, implying both more sites and more site types that, in turn, imply a commensurate increase in backhaul links.
On top of this macroscopic perspective are some intangible changes. Take technical complexity, for example. Each wireless generation is inherently more complex than its predecessor. Third-generation networks were near 10 times more complex than 2G in terms of processing demands and the number of call setup messages. Fourth-generation networks are similar and 5G will be the same again. As we sunset 2G and launch 5G, we are actually removing a network of complexity (1) and adding one of complexity (1,000). To understand how to tune such a complex and increasingly dense network becomes intellectually challenging.
Based on only these two points alone, excluding service complexity, new frequencies and other considerations, I think we can safely say network complexity is increasing. Whether it is beyond human capability to manage or not is probably a question that has no correct answer. However, if we imagine operating expense remains constant or even declines, then there is no doubt that something needs to be done. The only answer is to simplify and automate as much as we can.
The automation suite
Fortunately, there are a number of features that have been developed and are forthcoming that do help automate network operations. While typically not grouped together, nor even referred to as automation, it is in fact what they really do. They are customer experience management, self-organizing networks, analytics, software-defined networking and orchestration.
With different legacies, each has its own focus:
• CEM – managing the quality of experience of individual customers in real time.
• SON – access and core network tuning.
• Analytics – network and customer data mining.
• SDN – transport configuration.
• Orchestration – service deployment.
The value in looking at them as a group is that one can see their independent merits, as well as imagine combined functions.
For example, imagine that a CEM service determines that a customer’s throughput has decreased based on data from an analytics engine, and the root cause has been determined to be inappropriate BTS parameters and a poorly performing transport link. That CEM insight could move toward action by interfacing to SON to tune the RF parameters and to an SDN controller to select a new transport link. It may sound too good to be true; however, we have implemented practical customer-driven use cases where this CEM/SON link has been implemented and are now considering interfaces to SDN and other linkages.
Let’s try a more advanced network functions virtualization example. Suppose CEM detects that a marketing campaign threshold has been met and a new service needs to be launched. Perhaps CDMA voice usage has declined to a point where it may be appropriate to launch a voice-over-LTE campaign to help accelerate its sunset. That campaign could be launched by CEM through the NFV service orchestration function.
We can even look at linkages to systems developed for different purposes. For example, we have implemented a predictive operations service that draws on a self-learning analysis engine used in other industry sectors. By correlating a vast range of data, failures in the network or a service that are gradual and complex in nature can be detected and rectified. In this case, the automation linkages are analytics, artificial intelligence and SON.
As for the starting point, CEM is a good one because it prioritizes the customer first, and that’s not a bad thing. However, there are also other linkages that can be imagined independent of CEM, such as SON/SDN or orchestration/SDN.
One side effect of looking at automation tools in this manner, as alluded to earlier, is that you can easily imagine more integrated implementations. A common analytics engine that can draw on probes of all kinds and provides insights to different processing engines is perhaps the most obvious first step and one we have taken. Common data models, architectural principles and user interfaces could follow. It all comes from looking at the problem and system holistically.
Closing the loop – the benefits
Coming back to the foundational issue: Can we mask network complexity through usage of automation tools like those mentioned? The answer seems to be yes, through usage of them both independently and in combinations for increasingly sophisticated use cases. The key is having a holistic perspective.
And while not discussed, aside from enabling new use cases, a holistic perspective can also help bridge organizational silos within both operators and vendors; and that alone can lead toward long-term implementation and operational simplification.
Finally, while the original challenge was to simplify network operations, a side effect of all automation, when done well, is an ability to get things done faster and with fewer issues. That translates to happier customers and improved sales, which in turn helps network growth. And the loop closes.
Michael Murphy is CTO for Nokia Networks’ business in North America. In this capacity, Murphy is responsible for supporting customer engagements and influencing corporate strategy in product direction and roadmaps. Murphy joined Nokia in 2005 as Japan country manager, where he had profit and loss responsibility. He also led the formation of a senior team during the Softbank takeover of Vodafone. He then moved to the role of head of technology for the Asia-Pacific region, managing activities in Asia, the Middle East and Africa. In this function, his focus was roadmap planning for the region, LTE entry and development, and communicating the telecom vision with a focus on broadband profitability. Prior to joining Nokia, Murphy was the former head of W-CDMA development for Nortel Networks. Murphy has a master’s degree in mathematics from the University of Waterloo in Canada.
Editor’s Note: The RCR Wireless News Reality Check section is where C-level executives and advisory firms from across the mobile industry share unique insights and experiences.
Table of Contents
