5G User Plane Redundancy

We looked at the 5G Enhanced URLLC (eURLLC) earlier. One of the ways to improve reliability is to have redundancy in the user plane. This can use different approaches like: 

• Duplicating N3
• Adding a secondary gNB using Dual connectivity
• Introducing another UPF
• Two anchor UPFs

In fact they are all built on top of each other so you can decide how critical are your user plane redundancy needs. 

I came across this short video from Mpirical embedded below that covers this topic nicely. In case you want to refresh your 5G Core Network architecture, jump to our old tutorial here.

Related Posts:

• The 3G4G Blog: 5G Enhanced URLLC (eURLLC)
• The 3G4G Blog - Ultra Reliability: 5x9s (99.999%) in 3GPP Release-15 vs 6x9s (99.9999%) in 3GPP Release-16 
• The 3G4G Blog - Tutorial: Service Based Architecture (SBA) for 5G Core (5GC)
• 3G4G: 3GPP 5G Specifications
• 3G4G: 5G (IMT-2020) Wireless 

ITU Standardization Bureau on Machine Learning for 5G

Last year I blogged about Global ITU AI/ML 5G Challenge on the theme “How to apply ITU's ML architecture in 5G networks".  The grand challenge finale happened in December. All the recording and presentations are available here.

Back in October, Bilel Jamoussi from ITU presented a keynote to the 2020 IEEE 5G World Forum plenary session where he addressed the challenges of applying machine learning in networks, ITU’s ML toolkit, and ITU’s AI/ML in 5G Competition. IEEE Tv shared the presentation only in April so the competition part is a bit outdated. It does nevertheless an interesting 20 minute talk.

ITU Recommendation Y.3174, Framework for data handling to enable machine learning in future networks including IMT-2020 is available here.

Related Posts: 

• The 3G4G Blog: Artificial Intelligence (AI) / Machine Learning (ML) in 5G Challenge by ITU
• The 3G4G Blog: IMT-2020 (5G) Requirements 

3GPP RAN Plenary Update and Evolution towards 5G-Advanced

(click on image to enlarge)

ETSI recently held a webinar to provide a 3GPP RAN Plenary update by Wanshi Chen, Senior director of technology at Qualcomm Technologies, who was appointed as the RAN Chair not too long back. The webinar video is embedded below. The following is from the 3GPP summary of the webinar:

Wanshi Chen acknowledged that Release 17 - the third release of 5G specifications - has been under pressure due to COVID-19 restrictions, but despite making the move to e-meetings, he reported that the group’s experts have managed to ensure positive progress towards the freeze of the RAN1 physical layer specifications on schedule, by December 2021.

This is to be followed by the Stage 3 freeze (RAN2, RAN3 and RAN4) by March 2022 and the ASN.1 freeze and the performance specifications completion by September 2022 – On the timeline agreed back in December 2019.

This staggered timeline has been made achievable with careful planning and management, demonstrated to the webinar viewers via a complex planning schedule, with a slide showing the array of Plenary & WG meetings and Release landmarks - Interspersed with a series of planned periods of inactivity, to allow delegates some relief from 3GPP discussions.

Wanshi Chen noted that the efficiency of e-meetings has not been comparable with physical meetings, in terms of getting everything done. To compensate for that, the companies involved have planned two RAN1 meetings in 4Q21 and two meetings for each of the RAN working groups in the 1Q22. He observed: “We will monitor Release 17 RAN progress closely and take the necessary actions to make sure we can get the release completed on time.”

Release 18 Planning

Looking forward to Release 18 and the start of work on 5G-Advanced, Chen outlined the schedule for an online RAN workshop from June 28 – July 2, to define what will be in the release. The workshop will set the scene for email discussions about the endorsed topics for consideration. The work will culminate with Release 18 Package Approval, at the December 2021 Plenary (RAN#94).

The high-level objective of the workshop will be to gather company proposals in three areas:

• eMBB driven work;
• Non-eMBB driven functionality;
• Cross-functionality for both.

Wanshi Chen concluded that during the Release 18 planning process, some capacity must be kept in hand; keeping around 10% of WG effort in reserve, for workload management and to meet late, emerging critical needs from commercial deployments.

The following Q&A topics were covered, along with the time stamps:

• The effect of the pandemic and eMeeting management schedules and tools (19.25).
• Balance between commercial needs and societal needs, emergency services, energy efficiency, sustainability (21.20).
• The importance of the verticals in the second phase of 5G – With 5G-Advanced. How will this Rel-18 workshop compare in scale with the 5G Phoenix workshop in 2015? (23.00)
• The job of the Chair is to be impartial…but Wanshi guesses that Antennas, MiMo enh., Sidelink, Positioning, xR, AI machine learning…. could come up in Rel-18! (26.15)
• Will 5G-Advanced have a strong identity & support? (30.05)
• The potential for hybrid meetings – No clear answers yet, but we have learnt a lot in the past year.(34.35)
• The link between gathering new requirements and use cases in SA1 and RAN work and RAN1’s role in focusing these needs for radio work. (40.10)
• Software-ization of the RAN. Do you see more open RAN work coming to 3GPP? (44.18)
• Machine type communications and IoT – Where is IoT going in 3GPP RAN? (47.01)
• Some thoughts on Spectrum usage from a 3GPP point of view, is that difficult to fathom for non-experts? (52.00)
• Can Standards writing become more agile, less linear? (54.00)

If you want to get hold of the slides, you will have to register on BrightTALK here and then download from attachments.

Signals Research Group has a short summary of 3GPP RAN #91 electronic plenary held in late March. It is available to download after registration from here.

xoxoxoxoxoxo Updated later, 07 June 2021 oxoxoxoxoxoxox 

5G-Advanced logo is now available as shown above. Guidelines on how to use the logo is available on 3GPP here.

Related Posts:

• The 3G4G Blog: Initiative to Remove Non-inclusive terms from 3GPP Specifications
• The 3G4G Blog: ATIS Webinar on '5G Standards Developments in 3GPP Release 16 and Beyond'
• The 3G4G Blog: Anritsu Webinar on 'Evolution of 5G from 3GPP Rel-15 to Rel-17 and Testing Challenges'
• The 3G4G Blog: Spectrum for 5G NR beyond 52.6 GHz 

Anomaly Detection and other AI Algorithms in RAN Optimization

Yesterday I watched this very inspiring live chat that I would like to recommend to anyone who is interested in how machine learning techniques (aka "AI") can help to optimize and troubleshoot the Radio Access Network.

 [The real contents of in the video starts at approx. 42:00 min] 

My key takeaways from this fireside chat are: 

Verizon Wireless has enough data (100… 500 time series KPIs per cell) that they use to feed anomaly detection ML algorithms and this generates a huge number of alarms, but only a few actionable outputs. 

The “big elephant” (Nick Feamster) is to identify if these alarms indicating real problems that can and have to be fixed or if they just indicate a new behavior of e.g. a new handset or a SW version that was not present in the training phase of the ML algorithm and hence, its pattern is detected as a new “anomaly”. 

For Bryan Larish (Director Wireless AI Innovation, Verizon) the “big open problem” is “that it is not clear what the labels are” and “no standard training sets exist”. 

[For more details watch the video section between 52.00 min and 57:32 min and listen to Bryan’s experience!] 

In most cases Verizon seems to need subject matter experts to classify and label these anomaly alarms due to “the huge diversity” in data pattern. 

According to Bryan only for very few selected use cases it is possible to build an automated loop to fix the issue. Especially the root causes of radio interference are often mechanical or cabling issues that need manual work to get fixed. 

All in all it is my personal impression at the end of the session that anomaly detection is currently a bit overhyped and that the real challenges and problems to be resolved start after anomalies are detected.

Nevertheless, as Bryan summarizes: “ML is a very, very powerful tool.” 

However, strategically he seems not to see a lot of value in anomaly detection by itself, but rather: “Can we use machine learning (results) to change how we build networks in the future?”