Management, Orchestration & Automation Not Just Overhead Anymore

Despite a hiatus around 2017 as service providers struggled to create new OSS capabilities, this evolution is now culminating in 5G service management and orchestration built on the foundation of NFV network resource management and orchestration and on the re-architected Operations Support Systems (OSS) capabilities that are now integrally linked to Business – formerly billing – Support Systems (BSS).

With 5G’s Service Based Architecture (SBA) CSPs can finally begin to focus on service creation, enablement and orchestration to grow service revenues. Service orchestration now not only enables and assures CSP customers, it enables and assures service provider revenues.

5G SA Management will bring early Network Slicing revenues in 2020

One pre-eminent new source of 5G service revenues was expected to be 5G Network Slicing. Until recently, however, the network slicing market has been stalled largely for lack of cost effective dynamic management capabilities that assign and guarantee SLAs for a diverse set of slice types across a common pool (or pools) of physical network
resources.

Aggregating and Managing Dynamic Slice Assignment

5G now offers a way to aggregate and automate the dynamic assignment and control of Network Slicing for highly specialized apps and services, by treating multiple service flows as service instances that have similar characteristics from a network perspective and allocating them in real time to network slice instances that share common service parameters to create a solid business case for each slice.

To minimize the number of Network Slices required CSPs need to be able to characterize use cases and applications; and to classify those with similar performance and Class of Service requirements. These can now be specified in 5G as part of the template for a service. Although work is still continuing to define all the relevant parameters that can characterize service templates, some vendors are already offering a library, slice ‘store’ or ‘mall’ from which CSPs and their MVNO or Enterprise customers can select slice types and templates for instantiation as needed.

In addition going forward 5G slice management will allow service flows to be aggregated based on slice type/template characteristics so that multiple flows can run over a shared slice. Eventually service instances will be able to dynamically share Slice instances and – as NGMN originally envisioned – a common pool of underlying network resources without the high cost, inflexibility and loss of redundancy that ‘nailed up’ slices demand today.

Without such aggregation many IoT and uRLLC applications could be years away from individually generating profitable revenues. Similarly at the low end, IoT applications for groups of low level sensors or edge based latency tolerant services e.g. residential meter reading will be able to leverage 5G Slice management tools to share a common low priority network slice and deliver a viable CSP business case. For more details See Strategy Analytics July 2019 Report ‘Automated End to End Network Slice Management is a critical 5G ‘Service Enabler’ for profitable eMBB, uRLLC and mMTC’

5G Complexity demands large Scale Automation from RAN to Core

To deliver these flexible services however, 5G must manage enormous complexity in real time. The many diverse 5G use cases each demand precise performance across multiple hybrid domains from RAN to Core. 5G must manage and automate five critical sources of diversity – multi-technology, multi-band, multi-layer, massive MIMO and the multi-Architecture environment.

• Multi-Technology – 4G LTE and 5G NR must co-exist with 3G and 2G
• Multi-Band -5G will operate in multiple frequency bands from sub 1GHz and 3GHz- 6GHz to mmWave
• Multi-Layer – 5G small cells will offer densification or higher throughput ‘under’ a 4G coverage layer
• Massive MIMO – 5G NR is beam-based air interface and 3D beam planning/ optimization is very complex
• Multi-Architecture – Both RAN (CU and DU) and new SBA Core will co-exist with legacy in most cases

The massive scalability of 5G and the likely burstiness of 5G traffic both add additional dimensions to this complexity.
Yet CSPs must deliver the new 5G services with guaranteed performance at the same or lower cost per GB than they do over 4G, to protect their operating margins.

Operators can leverage automation and AI based on 5G management capabilities to:

1. Optimize 5G Efficiency across vast numbers of cells and smart antennas to:

• Automate HetNet configuration and RF planning with centralized SON
• Optimize RAN capacity and spectrum utilization in real time
• Manage small cells both alone and as clusters
• Configure and optimize RF carriers and backhaul flexibly in real-time
• Manage traffic loads based on Machine Learning for specific scenarios (e.g. cell edge
  peaks, variable usage by time of day)

2. Lower Total Cost of Operations (TCO).

In a 5G network with vast numbers of cell sites it is essential to reduce RAN maintenance and energy costs with network site optimization and automation.

3. Automate performance and operations processes end-to-end (E2E).

To ensure performance across network – all cells, carriers, antenna beams, backhaul links or edge computing elements etc. – must all be monitored and optimized in real time through end-to-end automation to ensure
performance.

SON and AI for 5G Management

SON is probably the most advanced automated Artificial Intelligence (AI) solution in service provider production networking today; and C-SON solutions now power managed services for LTE and 3G operators to accelerate CAPEX and OPEX savings as well as to optimize complex multivendor networks. In recent years SON has proven to be capable of running complete live production network operations, based on fully automated Machine Learning (ML) and ‘Closed loop’ Artificial Intelligence (AI).

First capture your dataset

Several industry experts have recently been heard to say that “because 5G is complex we should just apply AI and automate it”. But it is not quite that easy.

We are only just beginning to capture the datasets that represent new 5G traffic patterns under very diverse user scenarios. 5G is intentionally designed to handle such datasets – whether subscriber location and usage data in a central ‘data lake’ or network events and monitoring alerts in real time data stores at the edge –all 5G data will be accessible through Unified Data Management (UDM). Once 5G networks have been in operation long enough to capture significantly large datasets for multiple services, operators will be able to apply Machine Learning (ML)
and eventually produce Closed loop AI algorithms e.g. for RAN or backhaul load and congestion management. Unfortunately, the very bursty nature of 5G access is unlikely to fit the classic Poisson queuing models. We may need to develop new traffic models – e.g. for bursts of 8K video on downlink or for a stadium full of 5G handsets simultaneously uploading streaming video of an Olympic gold medal win. It may take many months or even years to capture all the relevant data and develop new models.

In parallel the complexity of RF planning for beam forming in 3D with Antenna Tilt and Massive MIMO options is one or two orders more complex that LTE RF planning!

ML and AI will help us to create these models – but we are still capturing the datasets with which to train the tools. It may be mid-2021 before we can begin to automate for 5G what we already do for LTE.

Zero Touch is coming, time to get started

Zero Touch management has been characterized as Service Driven, Lifecycle Centric, Intent Driven, Fluid, Intelligent, Cooperative, Autonomous and Decentralized – everything a service provider could ever want for a truly Service Driven Infrastructure.

Today Zero Touch still remains largely a long-term ideal that requires many new capabilities to be implemented in CSP networks before 5G can fully leverage it for service management and orchestration. Operators can however, start today by:

• Implementing End to End (E2E) service enablement, management and orchestration
• Evolving beyond MANO to service management and orchestration
• Automating Service Orchestration for select services
• Federating Orchestration across Service and Resource domains.

Network Automation will be Key to Operator 5G Success

As E2E network intelligence and automated processes are established across all network layers and throughout both network operations and service management, 5G will be able to scale to achieve its potential. These processes
can leverage comprehensive inputs from all network layers, elements and use cases – from eMBB, to mMTC and URLLC – and soon from dynamic Network Slicing and Multi-Access Edge Compute (MEC) services.

To deploy and operate high performance 5G networks a high degree of intelligent automation and eventually Zero Touch AI will be essential.

Contact:

Sue Rudd: email srudd@strategyanalytics.com