Posted by Scott Sumner on Tuesday, August 04, 2015 with No comments
Emergence of the instrumentation layer in real-time, intelligent networks.
5G networks will support a diverse array of applications, from 4k video streaming to safety-critical IoT, autonomous cars and factories, and hosted transactional applications that let us work from anywhere.
The competition for limited spectrum and network resources collides with increased complexity as networks layer-in virtualization to increase agility, manage network slices, and dynamically allocate capacity (both bandwidth and compute) where required.
SDN control is implicit to 5G, with the goal of harmonizing competing network demands, and permitting each application access to the resources and quality of service (QoS) it needs, while making most efficient use of infrastructure.
SDN controllers need to make real-time decisions to maintain balance in these ‘naturally unstable’ networks. Without constant adjustment, a static network will certainly be overwhelmed by the dynamic usage patterns in the network. Overprovisioning has long since been an affordable luxury.
A real-time view of the network state: all links, all bearers, all end-to-end services, each subscriber experience, and related subscriber ARPU, tier, policy entitlement, usage history, and location will need to be synthesized as feedback to SDN controllers striving to to use each resource as efficiently as possible, to dynamically optimize performance (DPO).
MNOs successfully moving to 5G are deploying big data analytics strategies in tandem with SDN networking initiatives to ensure their networks are ‘performance aware’, and optimize quality of experience (QoE) to differentiate their networks and drive them as efficiently as possible. This level of visibility also mitigates risk during the transition to SDN and NFV, as any change in the network can be predicted, then immediately quantified in terms of impact or improvement to subscriber QoE.
Complete Network State information can be captured with an instrumentation layer that brings data plane performance to the control plane via real-time analytics engines. The instrumentation layer combines network, connection, control and application layer KPIs and analyzed session metrics from physical and virtual infrastructure, hybrid SDN-based and legacy network elements, and both UE-centric, control and NFVI services.
Standards used for managing and monitoring the performance of backhaul networks are sufficient to define the tests, interfaces and data models required to establish such a network nervous system, but innovative methods to deploy such an instrumentation layer at near-zero cost and in the timescale of months pose the greatest challenge to the industry - one that needs to be solved as service assurance moves from a ‘nice to have’ to the foundation of user experience in 5G networks.
We see an opening for virtualized instrumentation as a facilitator to this vision of intelligent networks. When do you think this will emerge in real production networks? Let us know with a comment!