SON Needs to Smarten Up, to Succeed

Posted by Scott Sumner on Tuesday, August 09, 2016 with No comments


The time has come for self-optimizing networking (SON) to meet its neighbors. In fact, all the pieces of mobile networks need to start socializing—the cliques they have formed need to be broken up, and new forms of communication established to create a cohesive whole, as we move to the intelligent networks 5G will be.


SON controllers typically run autonomously, optimising radio transmission performance and mitigating interference, using metrics and network state information collected from the radios themselves. As software-defined networking (SDN) becomes mainstream in mobile core, the virtualized evolved packet core (vEPC) will need to coordinate with SON controllers through big data analytics to achieve a new level of user experience.

Consider the case where radios are fully optimised—from the SON controller's vantage point—but user experience suffers. Is the SON at fault? Is it doing its part to contribute to the best QoE, or is it impairing it? Centralised analytics and SDN will be able to sort this out, and put the SON network at the service of each end user's policies, most important applications, and their perception of what matters. For example, in a congested area users that pay a premium for a service should enjoy flawless performance. SON controllers would be instructed to optimise service quality around that user, potentially at the expense of other, less important subscribers or applications.

Some examples: 
  • When a user is engaging in VoLTE call, the SON may be instructed adjust radio properties, steer beams, or change backhaul paths to realise the lowest latency to the handset. 
  • A user streaming 4k video is relatively insensitive to delay, but needs significantly higher throughput. 
Only the combination of SON, with total performance visibility, analytics to process optimal scenarios, and an SDN controller to issue commands, can see and react to ensure the network delivers the best possible user experience.

This architecture has been proven out by some leading mobile network operators (MNOs), including Softbank, NTT, SK Telecom, Reliance, and T-Mobile. In each case, instrumenting the network proved to be a foundational step, as uniform network performance and quality of experience visibility over the SON, backhaul, transport and core networks is required for actionable analytics to converge on optimal network states. This is easier said than done in multi-vendor, multi-technology networks, many at different stages of virtualisation.

While virtualisation complicates the task of widespread visibility, it also holds the key to achieving it. Virtualised instrumentation that can be deployed on commercial off-the-shelf servers (COTS), in MNO data centers, and in mobile edge compute, permitting large scale instrumentation without requiring test and measurement hardware installation. Accedian has rolled out solutions using this approach at multiple Tier-1 MNOs this year, integrating network wide monitoring streams into big data analytics and SDN control systems within weeks.

Virtualised instrumentation offers another benefit here
since data is collected by virtualised controllers, it can be easily interfaced into virtualised database, orchestration and control platforms.

From there, analytics that correlate user experience with their context, policies, subscriptions and the network state, allow operators to see and control as never before, with the ability to add machine learning into the mix to create a fully intelligent, automated network.

Where do you think all this is heading, for the MNOs and the operations teams that hold it all together? It's a journey worth exploring and debating. The time is now!



Categories: , ,