Big Data Analytics vs. Smart Data Analytics

Mobile network operators have mountains of data they are starting to consult, crunch and even monetize to other parties interested in the behavior, location, and experience of subscribers. As mobile networks move to a service oriented architecture–consistent with intelligent, SDN, and virtualized infrastructure–access to the network state correlated with user metrics is also becoming foundational for network control, and ultimately, the feedback that will help providers optimize quality of experience (QoE).

The relationship between performance monitoring 
and customer experience is tightly tied to analytics.

Drowning in the ever wider, deeper data lake is becoming a real threat. From the perspective of intelligent network control that can dynamically optimize performance in response to real-time events, less data is more: quality over quantity prevails. Actionable, smart data analytics seeks to minimize the amount of data querying–and discard of imprecise or questionable metrics–to provide high resolving power to the key KPIs that can drive reliable, relevant control over the network, in real-time.

Without responsive, smart data analytics, the diversity of use cases and traffic sources justifying the shift to 5G networks will be difficult, if not impossible, to realize; real-time requirements make any information older than a minute unreliable for QoE control.

Following the principles of SDN, the instrumentation layer–the mediation layer capturing all layers of network and user metrics–must be programmable, and responsive, to the command of the control plane. Instead of collecting everything, all the time, intelligent instrumentation orchestration permits network control to ‘sense’ where additional information, granularity, localization, or multi-layer correlation is required, and dial-it on demand to realize responsive SON, dynamic backhaul optimization, and to preserve QoE under all network conditions.

Programmable, intelligent analytics rely on ubiquitous visibility 
where granularity can be increased on demand.

Smart data analytics must function at all layers, with the ability to prescriptively hone-in on relevant metrics from the network, control, session, protocol, application, and payload components. Analyzer platforms, SDN control, and machine learning platforms should be able to direct the instrumentation layer to provide the right feed of information at the right time, only as long as needed for optimal action to occur. Key aspects of smart data analytics that need to be developed: (1) real-time programmability, (2) ubiquitous metric / packet capture collection capabilities for all locations, network slices, services, sessions, users, policies, (3) standards-based, open interfaces for analytics aggregation and control, and (4) visualization methods that drive cost-efficient and agile adoption.

We can see these developments firming up, as we integrate open APIs, and all-layer monitoring into our product lines. Look to some major announcements in this regard as we head into 2016.