Analytics-Driven Automation Is Critical for Mobile Network Operators to Master 5G Complexity at Scale
By Andrew Colby, Head of 5G Strategy and Product Management, Guavus.
Operators that incorporate 3GPP-compliant data analytics into their networks from the outset can scale out and manage 5G deployments cost-effectively.
5G networks offer the promise of transforming wireless experiences with low-latency, gigabit data speeds delivered with ultra-reliability. However, these advances come at a price. Mobile network operators (MNOs) face unprecedented complexity as they begin to scale their 5G networks – complexity that can spiral beyond the abilities of humans using existing tools and only semi-automated workflows.
MNOs applying traditional processes to 5G network management will face challenges that place the economic returns of massive 5G investments at risk. However, those that integrate the Third-Generation Partnership Project’s (3GPP’s) Network Data Analytics Function (NWDAF) into their 5G Core can master the complexity of 5G by applying analytics-driven machine intelligence to network automation and service orchestration.
The 3GPP has defined a 5G Service-Based Architecture (SBA) that relies on network data analytics to continuously monitor network state across the 5G RAN/Core infrastructure, analyze this data in real time, and deliver statistical and predictive analytics outputs to Network Functions (NFs) and Application Functions (AFs) – which sit above the network layer – that these functions will consume to automate 5G network and service operations.
5G Services Go Cloud Native
Standalone 5G Core networks will operate on cloud-native platforms using the same technologies that powerhouses like Amazon, Microsoft, and Google use to deliver cloud-based services at massive scale. Cloud-native infrastructure is disaggregated, virtualized and software-defined, enabling MNOs to rapidly conceive, develop and deploy a new generation of 5G services by employing the same state-of-the-art DevOps methodologies that IT organizations have adopted to manage the operations of today’s digital service providers. Real-time data analytics is a critical component of the DevOps CI/CD (Continuous Integration/Continuous Deployment) lifecyle that delivers constant feedback about operational state that is utilized to drive (automatically or manually) any system changes or modifications needed to ensure that service operations meets SLAs.
4G/LTE Tools and Practices Not Well-Suited for 5G
5G RAN/Core infrastructure will support highly dense networks with thousands of small cells packed into relatively small geographic areas and massive scale IoT connectivity delivering 5G service to millions of smart devices. In addition to scale, 5G networks will be far more dynamic than 4G/LTE networks, and so it will be even more critical to continuously monitor the state of the network and the behavior of connected devices to track network performance, detect faults and take any actions required to ensure service continuity.
Existing OSS tools and operational practices were not conceived with these considerations in mind. While 4G/LTE operators have had some success employing analytics and machine learning in existing networks, these efforts have required a customized “bolt-on” approach, due to a lack of standards for data analytics and a system architecture that was optimized for network operations centers staffed by humans. With 5G, the 3GPP is ensuring that analytics is not an afterthought, and that analytics-driven automation is built into the system architecture.
Data Analytics Standards Facilitate 5G Multi-Vendor Interoperability
The 3GPP’s 5G data analytics standards are also critical to facilitate the integration of 5G RAN/Core system components from multiple vendors. For example, NWDAF specifies standard data types, formats, data collection APIs and standard data outputs and APIs for analytics processing. This is important because 5G cloudification and the scope of 5G services to be deployed will foster the growth of a 5G supplier ecosystem which will be far more diverse than the existing 4G/LTE ecosystem, which today is dominated by a relatively small number of suppliers.
NWDAF: 3GPP Standard for Network Data Analytics
NWDAF represents the mobile industry’s first attempt to standardize the function of analytics in the mobile core network. NWDAF incorporates standard interfaces for collecting different types of data from certain 5G Core NFs and applies the results of analytics processing to inform the operation of other NFs, applying machine intelligence to network automation and service orchestration.
A key problem that NWDAF addresses is data normalization across dissimilar interfaces and data formats in multi-vendor networks. This problem has historically made data collection, aggregation, integration, and analysis from different suppliers’ equipment difficult and time-consuming, resulting in a problematic ROI for many data analytics projects. That situation is now changed. A key part of solving the data normalization problem also involves specifying data semantics for the statistical and predictive analytics outputs that are delivered to NFs and AFs in the 5G Core.
Operationally, NWDAF runs as an NF in the 5G Core network, and once it is deployed, NWDAF registers its Analytics IDs with the Network Repository Function (NRF) – a centralized repository for all of the 5G NFs – and discovers all NFs it needs to communicate with. From there, OAM systems can gather operational intelligence quickly and easily from NWDAF. The data output by NWDAF is easily utilized by the consumer NFs in the 5G Core, or by 5G Management Data Analtyic Functions (MDAFs) that can also integrate data from an array of other sources spanning the 5G Core, 5G RAN, external data networks, the underlying transport network, and the mobile edge network (see figure). All this takes place without the need for any data reformatting or normalization.
Today, most networks billed as “5G” comply with the 3GPP’s 5G New Radio (NR) standards for the RF portions of the network but with the radios connected via a 4G/LTE core. While this is an expedient way to quickly build out 5G NR sites, MNOs won’t realize the full business potential of 5G until they are operating in 5G Standalone (SA) mode, with the radios connected via a 5G Core. The 3GPP has designed the new, cloud-native 5G Core has to support the network-wide data collection and analytics processing needed to derive the critical statistical and predictive insights that will power 5G network automation and service orchesteration.
NWDAF Defines Standard Use Cases
In addition to standard data inputs, outputs and APIs, the 3GPP has also defined a set of standard NWDAF use cases, which specify the source NFs, the statistical or predictive analytics outputs, and the consuming NFs for each particular use case. The following are several examples. For a complete list of standard NWDAF use cases in 3GPP Release 16, see the sidebar, “3GPP-Defined NWDAF Use Cases.”
- Network slicing. This is a capability in 5G networks that’s analogous to virtual LANs (VLANs) in IP networks, creating logical segmentation between end points or applications over a common physical network infrastructure. In a 5G environment with tens or hundreds of network slices, it will be challenging to determine which network slice can provide the best service to a given device. To address this, one use case defined for NWDAF is identifying and predicting the load per individual NF and for each network slice instance. The Network Slice Selection Function (NSSF) can use this information from the NWDAF to help it determine to which network slice a newly registered device (also known as “UE,” or “user equipment”) should be assigned.
- Session load balancing. Similarly, an Access and Mobility Management Function (AMF), which manages subscribers’ access to the network, might request specific intelligence on the load level of several Session Management Function (SMF) instances in order to assign a UE to the SMF best able to serve it.
- Policy decision making. A policy check allows the PCF to apply appropriate policies to the device connection based on the current state of the network. The PCF can use information from the NWDAF for the observed service experience of a device to determine if the application SLA is being satisfied, and if not, what QoS parameters should be applied for the service.
Don’t Defer NWDAF Until Later
MNOs migrating networks from 4G/LTE to 5G face unprecedented changes in scale and complexity that will impact their ability to meet operating cost and performance targets employing traditional operations tools and processes. It is imperative that MNOs factor network automation and service orchestration into their plans from the outset, which will require adopting the 3GPP’s standards-based approach to 5G network data analytics.
MNOs can’t afford to defer NWDAF and 5G data analytics until “later” because without NWDAF they lose the ability to automate network operations using the mechanisms designed into the 5G Core’s SBA. The NWDAF standard can be implemented and deployed now, and as described above, it also helps facilitate the integration of multi-vendor 5G RAN/Core infrastructure. MNOs that wait to deploy 5G data analytics until after building out their 5G network risk seeing the cost of 5G operations quickly spiral out of control as a function of scale. NWDAF is a critical, non-optional 5G Core NF that operators need to build into their network from the start.
Management Data Analytics and RAN Analytics
The 3GPP 5G SBA also defines Management Data Analytics Functions (MDAFs) which support the collection and analysis of OAM data for a broad range of management capabilities, including automated service assurance, fault management, performance management, and provisioning. The intent is to define a standard “form factor” that will streamline the development OAM data analytics in 5G systems, which will be particularly helpful in multi-vendor environments.
Real-time analytics will be critical for automating the operation of highly complex 5G NR networks in the baseband processing and RF domains. An example of a 5G standard in this area is the Open RAN Alliance’s specification for the RAN Intelligent Controller. While this specification has been developed outside of the 3GPP, adoption of this standard will be critical for the successful operation of multi-vendor 5G Open RAN deployments.
3GPP-Defined NWDAF Use Cases
The 3GPP has defined 10 specific NWDAF analytics use cases in Release 16 of the NWDAF standard:
- Network slice instance load level computation and prediction
- Service experience computation and prediction for an application or UE group
- Load analytics information and prediction for a specific NF
- Application service experience computation and prediction
- Network performance computation and prediction
- UE mobility analytics and expected behavior prediction
- UE abnormal behavior/anomaly detection
- UE communication analytics and pattern prediction
- Congestion information – current and predicted for a specific location
- Quality of service (QoS) sustainability – reporting and predicting QoS change
Andrew Colby is Head of 5G Strategy and Product Management at Guavus, a pioneer in AI-based analytics for communications service providers. As a member of the Guavus Office of the CTO, Andrew leads initiatives with customers to identify ways to apply analytics to improve and transform their operations and customer experience. He has worked in the areas of telecom and IP networking, operational support systems, and data analytics, for more than 30 years.
Photo by Joshua Sortino on Unsplash