09/16/2021 | Press release | Distributed by Public on 09/16/2021 12:44
By Rick Fulwiler and John English
Network Data Analytics Function (NWDAF) is a new 3GPP-defined analytics function in the 5G core network responsible for providing network analysis information to network functions and application functions upon request. It collects and analyzes data from user equipment, network functions, application functions, and operations and management (OAM) systems across the 5G core, cloud, edge networks, and radio access networks (RANs).
Equipped with specific automation techniques and AI algorithms, NWDAF uses data from network functions and application functions to provide near real-time analysis back to the network and application functions. NWDAF also provides predictive analysis that helps to proactively manage these functions with less human intervention.
NWDAF is one of several data analytics capabilities that comprise the analytics frameworks for the 5G core network. According to the 5G-PPP Architecture Working Group View on 5G Architecture Version 3.0, this analytics framework considers data analytics capabilities at various layers and introduces data analytics functions (DAFs) into the following:
In a heterogeneous environment such as 5G, it becomes important for each function to continuously evolve and provide better services to end-user devices. This can be achieved only if the functions in the 5G core act proactively and make real-time decisions, which in turn can provide necessary services to 5G consumers without delays. With a cloudified 5G network that is expected to support a huge number of devices, from smart phones to a myriad of IoT applications and devices, each network function must use automation to evolve toward self-optimization.
For example, DAFs could send key performance indicators to a policy control function if GTP packet delay exceeds a specific service level agreement governing quality of service for a given slice. Each logical data analytics module is implemented as multiple instances for different use cases and purposes. For instance, the big data/MDAF module could be implemented as multiple instances at different domain levels, such as RAN data analytics and VNF data analytics cross/intradomain. This framework allows for a dedicated data analytic module design at different layers, also enabling cross-layer optimization. Figure 1 depicts the overall integrated data analytics framework.
Figure 1: NWDAF Architecture
Phased Approach to NWDAF
As with DAFs in general, NWDAF isn't an off-the-shelf solution like a typical network equipment manufacturer (NEM) virtual function in 5G stand-alone (SA) architecture. It will have the core ability to acquire packets from the service-based interface and feed data to external functions and data repositories, but the analytics engine within will be highly programable based on whatever use case a mobile operator wants NWDAF to execute. This is a very important concept, because the mobile operator will still want very tight control of the 5G SA network. Engineering and operations teams will not, at this time, turn network control over to an AI engine without very specific human checks and balances in place.
Instead, mobile operators will likely first spend time learning how to best use NWDAF analysis to improve the operational efficiency of the 5G SA network. Communication service providers (CSPs) will need to learn how to tune the AI algorithms within NWDAF to fit their business parameters and their level of comfort with using automated network configurations based on dynamic network or service conditions.
What 5G Services Benefit from NWDAF?
All 5G services will benefit from NWDAF, but low-latency network slices will need the most immediate real-time adjustments with live network operation. Early field trials and rollouts are projected to be Open RAN-related, with the core primary focus on low-latency optimization. Soon to follow are user experience analytics, especially in the areas of cloud gaming and virtual reality. Network slicing is still slow in coming to the carriers, but with it we expect that affected user experience detection/prediction cases will be important.
The key aspect is for CSPs to use the highest resolution of data-something that NETSCOUT provides off the wire, along with a flexible use case-driven AI/machine learning engine that can harness smart data into actionable results.
Figure 2: Automated Intelligence in Omnis NWDAF
As 5G network functions increasingly use automation, many expect NWDAF will become essential to the operation of 5G networks in the near term. The implementation of closed-loop orchestration is not gated completely by the NWDAF specification. Automation in 5G networks will likely happen organically as well. Even today in non-5G networks, NETSCOUT data is a critical component for cell congestion and policy control.
Significant Drivers for NWDAF
Automation and closed-loop orchestration will likely happen with or without conformance to an industry specification. Large carriers that demand conformance to NWDAF specifications will likely be significant drivers for the adoption.
Vendors have also developed and delivered non-NWDAF solutions for network automation over the past few years. For example, NETSCOUT's Adaptive Service Intelligence Smart Data has been used in many forms of automation that until now have not been given specific APIs and/or specifications.
The DAF standards are currently somewhat loosely defined, which is a good thing for NEMS and carriers getting started with closed-loop implementations. What is needed is not more definition in the standards, but rather real-world implementation of closed-loop orchestration use cases.
NWDAF is likely be a work in progress for the foreseeable future, as CSPs roll it out in multiple phases to give engineering and operations teams time to learn and gain confidence in how to best use this technology. Having a vendor-agnostic partner with cloud-optimized smart data will be indispensable in a multivendor implementation of NWDAF.
Learn more about the importance of 5G analytics and network performance