Showing posts with label Machine Learning. Show all posts
Showing posts with label Machine Learning. Show all posts

Tuesday, 2 February 2021

NWDAF in 3GPP Release-16 and Release-17

We looked at Network Data Analytics Function, NWDAF, in detail here. While the 3GPP Release-16 work just starting back then, we have now completed Rel-16 and looking at Release 17. 

The 5G Core (5GC) supports the application of analytics to provide Intelligent Automation of the network, In Rel-16 the set of use cases that are proposed for the NWDAF has been widely expanded. 

In an earlier post, we looked at the ATIS webinar discussing Release-16 & forthcoming features in Rel-17. Puneet Jain, Director of Technical Standards at Intel and 3GPP SA2 Chairman talked briefly about NWDAF. The following is from his talk:

Release-16 provides support for Network Automation and Data Analytics.  Network Data Analytics Function (NWDAF) was defined to provide analytics to 5G Core Network Functions (NFs) and to O&M. It consists of several services that were defined in 3GPP Rel-16 and work is now going in Release 17 to further extend them. 

In release 16 Slice load level related network data analytics and observed service experience related network data analytics were defined. NF load analytics as well Network Performance analytics was also specified. NWDAF provides either statistics or prediction on the load communication and mobility performance in the area of interest. 

Other thing was about the UE related analytics which includes UE mobility analytics, UE communication analytics, Expected UE behavior parameter, Related network data analytics and abnormal behavior related network data analytics.

The NWDAF can also provide user data congestion related analytics. This can be done by one time reporting or continuous reporting in the form of statistics or prediction or both to any other network function. 

QoS sustainability analytics, this is where the consumer of QoS sustainability analytics may request NWDAF analytics information regarding the QoS change statistic for a specific period in the past in a certain area or the likelihood of QoS change for a specific period in future, in certain areas. 

In Release 17, studies are ongoing for network automation phase 2. This includes some leftover from Release 16 such as UE driven analytics, how to ensure that slice SLA is guaranteed and then also new functionality is being discussed that includes things like support for multiple NWDAF instance in one PLMN including hierarchies, how to enable real-time or near-real-time NWDAF communications, how to enable NWDAF assisted user pane optimization and last which is very interesting is about interaction between NWDAF and AI model and training service owned by the operator.

This article on TM Forum talks about NWDAF deployment challenges and recommendations:

To deploy NWDAF, CSPs may encounter these challenges:

  • Some network function vendors may not be standards compliant or have interfaces to provide data or receive analytics services.
  • Integrating NWDAF with existing analytics applications until a 4G network is deployed is crucial as aggregated network data is needed to make decisions for centralized analytics use cases.
  • Many CSPs have different analytics nodes deployed for various use cases like revenue assurance, subscriber/marketing analytics and subscriber experience/network management. Making these all integrated into one analytics node also serving NWDAF use cases is key to deriving better insights and value out of network data.
  • Ensuring the analytics function deployed is integrated to derive value (e.g., with orchestrator for network automation, BI tools/any UI/email/notification apps for reporting).

Here are some ways you can overcome these challenges and deploy efficient next-generation analytics with NWDAF:

  • Mandate a distributed architecture for analytics too, this reduces network bandwidth overhead due to analytics and helps real-time use cases by design.
  • Ensure RFPs and your chosen vendors for network functions have, or plan to have, NWDAF support for collecting and receiving analytics services.
  • Look for carrier-grade analytics solutions with five nines SLAs.
  • Choose modular analytics systems that can accommodate multiple use cases including NWDAF as apps and support quick development.
  • Resource-efficient solutions are critical for on-premise or cloud as they can decrease expenses considerably.
  • Storage comes with a cost, store more processed smart data and not more raw big data unless mandated by law.
  • In designing an analytics use case, get opinions from both telco and analytics experts, or ideally an expert in both, as they are viewed from different worlds and are evolving a lot.

This is such an important topic that you will hear more about it on this blog and elsewhere.

Related Posts:

Saturday, 1 August 2020

Artificial Intelligence (AI) / Machine Learning (ML) in 5G Challenge by ITU


ITU is conducting a global ITU AI/ML 5G Challenge on the theme “How to apply ITU's ML architecture in 5G networks". If you don't know the difference between AI & ML, this picture from the old blog post may help.


The ITU website says:

Artificial Intelligence (AI) will be the dominant technology of the future and will impact every corner of society. In particular, AI / ML (machine learning) will shape how communication networks, a lifeline of our society, will be run. Many companies in the ICT sector are exploring how to make best use of AI/ML. ITU has been at the forefront of this endeavour exploring how to best apply AI/ML in future networks including 5G networks. The time is therefore right to bring together the technical community and stakeholders to brainstorm, innovate and solve relevant problems in 5G using AI/ML. Building on its standards work, ITU is conducting a global ITU AI/ML 5G Challenge on the theme “How to apply ITU's ML architecture in 5G networks". 

Participants will be able to solve real world problems, based on standardized technologies developed for ML in 5G networks. Teams will be required to enable, create, train and deploy ML models (such that participants will acquire hands-on experience in AI/ML in areas relevant to 5G). Participation is open to ITU Member States, Sector Members, Associates and Aca​demic Institutions and to any individual from a country that is a member of ITU. ​

There are also some cash prizes, etc. There are various topics with presentation slides & recordings freely available. 

I found the slides from ITU AI/ML in 5G Challenge —”Machine Learning for Wireless LANs + Japan Challenge Introduction” (link) very interesting. There are many other topics including AR / VR / XR, etc, as well.

Have a look at the ITU website here.


Related Posts:

Wednesday, 6 February 2019

AI in 5G – the why and how

IET recently held the 6th Annual 5G conference bringing together key players in the 5G world. You can watch the videos for that event here (not all have been uploaded at the time of writing this post).

We reached out to Dr. Yue Wang to share her presentation with us and she has kindly done so. The presentation and video are embedded below.






Related posts:

Monday, 13 August 2018

Telefonica: Big Data, Machine Learning (ML) and Artificial Intelligence (AI) to Connect the Unconnected


Earlier, I wrote a detailed post on how Telefonica was on a mission to connect 100 Million Unconnected with their 'Internet para todos' initiative. This video below is a good advert of what Telefinica is trying to achieve in Latin America


I recently came across a LinkedIn post on how Telef√≥nica uses AI / ML to connect the unconnected by Patrick Lopez, VP Networks Innovation @ Telefonica. It was no brainer that this needs to be shared.



In his post, Patrick mentions the following:

To deliver internet in these environments in a sustainable manner, it is necessary to increase efficiency through systematic cost reduction, investment optimization and targeted deployments.

Systematic optimization necessitates continuous measurement of the financial, operational, technological and organizational data sets.

1. Finding the unconnected


The first challenge the team had to tackle was to understand how many unconnected there are and where. The data set was scarce and incomplete, census was old and population had much mobility. In this case, the team used high definition satellite imagery at the scale of the country and used neural network models, coupled with census data as training. Implementing visual machine learning algorithms, the model literally counted each house and each settlement at the scale of the country. The model was then enriched with crossed reference coverage data from regulatory source, as well as Telefonica proprietary data set consisting of geolocalized data sessions and deployment maps. The result is a model with a visual representation, providing a map of the population dispersion, with superimposed coverage polygons, allowing to count and localize the unconnected populations with good accuracy (95% of the population with less than 3% false positive and less than 240 meters deviation in the location of antennas).


2. Optimizing transport



Transport networks are the most expensive part of deploying connectivity to remote areas. Optimizing transport route has a huge impact on the sustainability of a network. This is why the team selected this task as the next challenge to tackle.

The team started with adding road and infrastructure data to the model form public sources, and used graph generation to cluster population settlements. Graph analysis (shortest path, Steiner tree) yielded population density-optimized transport routes.


3. AI to optimize network operations


To connect very remote zones, optimizing operations and minimizing maintenance and upgrade is key to a sustainable operational model. This line of work is probably the most ambitious for the team. When it can take 3 hours by plane and 4 days by boat to reach some locations, being able to make sure you can detect, or better, predict if / when you need to perform maintenance on your infrastructure. Equally important is how your devise your routes so that you are as efficient as possible. In this case, the team built a neural network trained with historical failure analysis and fed with network metrics to provide a model capable of supervising the network health in an automated manner, with prediction of possible failure and optimized maintenance route.

I think that the type of data driven approach to complex problem solving demonstrated in this project is the key to network operators' sustainability in the future. It is not only a rural problem, it is necessary to increase efficiency and optimize deployment and operations to keep decreasing the costs.


Finally, its worth mentioning again that I am helping CW (Cambridge Wireless) organise their annual CW TEC conference on the topic 'The inevitable automation of Next Generation Networks'. There are some good speakers and we will have similar topics covered from different angles, using some other interesting approaches. The fees are very reasonable so please join if you can.

Related posts:

Sunday, 29 July 2018

Automating the 5G Core using Machine Learning and Data Analytics

One of the new entities introduced by 3GPP in the 5G Core SBA (see tutorial here) is Network Data Analytics Function, NWDAF.
3GPP TR 23.791: Study of Enablers for Network Automation for 5G (Release 16) describes the following 5G Network Architecture Assumptions:

1 The NWDAF (Network Data Analytics Function) as defined in TS 23.503 is used for data collection and data analytics in centralized manner. An NWDAF may be used for analytics for one or more Network Slice.
2 For instances where certain analytics can be performed by a 5GS NF independently, a NWDAF instance specific to that analytic maybe collocated with the 5GS NF. The data utilized by the 5GS NF as input to analytics in this case should also be made available to allow for the centralized NWDAF deployment option.
3 5GS Network Functions and OAM decide how to use the data analytics provided by NWDAF to improve the network performance.
4 NWDAF utilizes the existing service based interfaces to communicate with other 5GC Network Functions and OAM.
5 A 5GC NF may expose the result of the data analytics to any consumer NF utilizing a service based interface.
6 The interactions between NF(s) and the NWDAF take place in the local PLMN (the reporting NF and the NWDAF belong to the same PLMN).
7 Solutions shall neither assume NWDAF knowledge about NF application logic. The NWDAF may use subscription data but only for statistical purpose.

Picture SourceApplication of Data Mining in the 5G Network Architecture by Alexandros Kaloxylos

Continuing from 3GPP TR 23.791:

The NWDAF may serve use cases belonging to one or several domains, e.g. QoS, traffic steering, dimensioning, security.
The input data of the NWDAF may come from multiple sources, and the resulting actions undertaken by the consuming NF or AF may concern several domains (e.g. Mobility management, Session Management, QoS management, Application layer, Security management, NF life cycle management).
Use case descriptions should include the following aspects:
1. General characteristics (domain: performance, QoS, resilience, security; time scale).
2. Nature of input data (e.g. logs, KPI, events).
3. Types of NF consuming the NWDAF output data, how data is conveyed and nature of consumed analytics.
4. Output data.
5. Possible examples of actions undertaken by the consuming NF or AF, resulting from these analytics.
6. Benefits, e.g. revenue, resource saving, QoE, service assurance, reputation.

Picture SourceApplication of Data Mining in the 5G Network Architecture by Alexandros Kaloxylos

3GPP TS 23.501 V15.2.0 (2018-06) Section 6.2.18 says:

NWDAF represents operator managed network analytics logical function. NWDAF provides slice specific network data analytics to a NF. NWDAF provides network analytics information (i.e., load level information) to a NF on a network slice instance level and the NWDAF is not required to be aware of the current subscribers using the slice. NWDAF notifies slice specific network status analytic information to the NFs that are subscribed to it. NF may collect directly slice specific network status analytic information from NWDAF. This information is not subscriber specific.

In this Release of the specification, both PCF and NSSF are consumers of network analytics. The PCF may use that data in its policy decisions. NSSF may use the load level information provided by NWDAF for slice selection.

NOTE 1: NWDAF functionality beyond its support for Nnwdaf is out of scope of 3GPP.
NOTE 2: NWDAF functionality for non-slice-specific analytics information is not supported in this Release of the specification.

3GPP Release-16 is focusing on 5G Expansion and 5G Efficiency, SON and Big Data are part of 5G Efficiency.
Light Reading Artificial Intelligence and Machine Learning section has a news item on this topic from Layer123's Zero Touch & Carrier Automation Congress:

The 3GPP standards group is developing a machine learning function that could allow 5G operators to monitor the status of a network slice or third-party application performance.

The network data analytics function (NWDAF) forms a part of the 3GPP's 5G standardization efforts and could become a central point for analytics in the 5G core network, said Serge Manning, a senior technology strategist at Sprint Corp.

Speaking here in Madrid, Manning said the NWDAF was still in the "early stages" of standardization but could become "an interesting place for innovation."

The 3rd Generation Partnership Project (3GPP) froze the specifications for a 5G new radio standard at the end of 2017 and is due to freeze another set of 5G specifications, covering some of the core network and non-radio features, in June this year as part of its "Release 15" update.

Manning says that Release 15 considers the network slice selection function (NSSF) and the policy control function (PCF) as potential "consumers" of the NWDAF. "Anything else is open to being a consumer," he says. "We have things like monitoring the status of the load of a network slice, or looking at the behavior of mobile devices if you wanted to make adjustments. You could also look at application performance."

In principle, the NWDAF would be able to make use of any data in the core network. The 3GPP does not plan on standardizing the algorithms that will be used but rather the types of raw information the NWDAF will examine. The format of the analytics information that it produces might also be standardized, says Manning.

Such technical developments might help operators to provide network slices more dynamically on their future 5G networks.

Generally seen as one of the most game-changing aspects of 5G, the technique of network slicing would essentially allow an operator to provide a number of virtual network services over the same physical infrastructure.

For example, an operator could provide very high-speed connectivity for mobile gaming over one slice and a low-latency service for factory automation on another -- both reliant on the same underlying hardware.

However, there is concern that without greater automation operators will have less freedom to innovate through network slicing. "If operators don't automate they will be providing capacity-based slices that are relatively large and static and undifferentiated and certainly not on a per-customer basis," says Caroline Chappell, an analyst with Analysys Mason .

In a Madrid presentation, Chappell said that more granular slicing would require "highly agile end-to-end automation" that takes advantage of progress on software-defined networking and network functions virtualization.

"Slices could be very dynamic and perhaps last for only five minutes," she says. "In the very long term, applications could create their own slices."

Despite the talk of standardization, and signs of good progress within the 3GPP, concern emerged this week in Madrid that standards bodies are not moving quickly enough to address operators' needs.

Caroline Chappell's talk is available here whereas Serge Manning's talk is embedded below:



I am helping CW organise the annual CW TEC conference on the topic The inevitable automation of Next Generation Networks
Communications networks are perhaps the most complex machines on the planet. They use vast amounts of hardware, rely on complex software, and are physically distributed over land, underwater, and in orbit. They increasingly provide essential services that underpin almost every aspect of life. Managing networks and optimising their performance is a vast challenge, and will become many times harder with the advent of 5G. The 4th Annual CW Technology Conference will explore this challenge and how Machine Learning and AI may be applied to build more reliable, secure and better performing networks.

Is the AI community aware of the challenges facing network providers? Are the network operators and providers aware of how the very latest developments in AI may provide solutions? The conference will aim to bridge the gap between AI/ML and communications network communities, making each more aware of the nature and scale of the problems and the potential solutions.

I am hoping to see some of this blog readers at the conference. Looking forward to learning more on this topic amongst others for network automation.

Related Post: