An ambitious programme geared to create a radically new architecture for the UK’s internet and telecommunications infrastructure
tech talks 2023...
A new series of short technical talks from PhD students and Researchers on the project..
Creating conditions for effective technology implementation.
Eleanor Toye Scott
Research Associate - Judge Business School, Cambridge
7th March 2023, from 13.00 - 13.30
Digital transformation, through the adoption of artificial intelligence (AI) and machine learning (ML), offers organizations expanding opportunities for greater control and efficiency and more timely and accurate results, but at the same time brings escalating emergent risks. This talk will focus on recent results from our research on technology and organizations for NG-CDI, including our findings at BT. I will draw out two main themes for discussion: firstly, the role of organizational factors in ensuring preparedness for technology deployment, and secondly, some of the organizational challenges raised by the technologies themselves.
​
Dr Eleanor Toye Scott is a Research Associate at the Department of Computer Science and Technology, University of Cambridge, and until very recently at the Judge Business School, University of Cambridge. Her PhD was in Experimental Psychology, and she has worked in academia and industry at the intersection of Human Computer Interaction and Organisational Behaviour. Her research interests include design and integration of complex technologies into organisations, oversight and risk management of automated and AI systems over different timescales and stakeholder engagement with technological, social and organisational change.
Transforming the management of complex infrastructures using multilayer network modelling
Manuel Herrera
Senior Research Associate, University of Cambridge
14th March 2023, from 13.00 - 13.30
The reliable operation of critical infrastructures requires sophisticated methods to investigate and understand their behaviours from a variety of different business and operational viewpoints.
Powerful multi-layer network representations of complex infrastructures provide a range of views and methods that allow the business to navigate a number of interrelated viewpoints. This enables much more holistic decisions to be made about operations on a number of timescales. It helps investigate possible future designs and responses to different operating conditions.
One and the same data-driven representation can be used to analyse the criticality of network nodes (routers), or dependence on cables and links, either in quasi-real time or as a long-term trend. Criticality can be determined by the likelihood of a detriment to service, or of a broader service or business risk. This helps the business design the best course of action: from determining which interventions will be most effective, to how to design more robust network structures in the first place.
​
Dr Manuel Herrera is a Senior Research Associate (SRA) in distributed intelligent systems at the University of Cambridge (UK). His research focuses on predictive analytics and statistical learning in complex systems for the optimal operation and management of smart and resilient critical infrastructure, with a particular emphasis on urban water systems. He also has considerable experience in telecommunication, transport, and maritime sectors through his involvement in projects funded by Industry, the UK government and research councils. Dr Herrera is a fellow of the Royal Statistical Society (RSS) and committee member of the RSS Special Interest Group in statistical engineering. He is also the recipient of the Frank Hansford-Miller fellowship 2021 in applied statistics, awarded by the WA Branch of the Statistical Society of Australia; and the recipient of the Excellence in Research award 2022, granted by the Institute for Manufacturing, Dept. of Engineering, at the University of Cambridge.
Taming network complexity with model driven Intents
Paul Alcock
PhD Student, Lancaster University
21st March 2023, from 13.00 - 13.30
Intent-based Networking seeks to realise automated network management through a contextual understanding of network infrastructure and abstract service goals. Practical implementations of intent remain limited however, with much of the domain targeting specialised use-cases and optimisation. Semantic Web technologies aim to support automated reasoning strategies through the use of contextual ontologies/knowledge graphs and logic rules. In this talk, I will present an initial approach towards intent-based networking, which aims to exploit the contextual reasoning capacity of semantic technologies for automated network operations.
​
Paul is a second-year PhD Student working at Lancaster University on intent-based networking for the NG-CDI project. His research considers the use of semantic web technologies to facilitate intent-driven decision-making for autonomic network management
​
Optimising Costs and Performance of 5G Open-RANs using Heuristic
Load Balancing Algorithms
Esmaeil Amiri
PhD student, University of Surrey
4th April 2023, from 13.00 - 13.30
Reliable service and efficiency of OpenRAN 5G networks depends on optimal load balancing between network resources. Balancing traffic loads between Centralised Units (CUs) and mid-haul transmission affects bandwidth requirements and packet delay. The problem is NP-hard so no solution can be found on practical timescales. Instead, a heuristic algorithm is proposed.
The analysis of different centralization levels indicates that using multi-CUs could reduce the total bandwidth usage by up to 20%. This is a significant improvement in resource utilization, which translates to network cost savings. Additionally, the use of multipath routing can improve the result of load balancing between mid-haul links, whilst increasing bandwidth usage, leading to improved network performance.
​
Esmaeil has a background in electronic engineering, and he is currently pursuing his Ph.D. at the Institute for Communication Systems at the University of Surrey. His research interests are primarily focused on network architecture, optimization, and reinforcement learning, with a specific focus on Open-RAN (O-RAN) technology.
AI-Enabled Network Design for Virtualized RAN
Xiaoyang Wang
Lecturer, University of Exeter
18th April 2023, from 13.00 - 13.30
Mobile network operators envision a future in which upgrading network infrastructure is as simple as upgrading a user device. Virtualized RAN (vRAN) enables this evolution, with the potential to transform the way mobile networks are designed, deployed, and managed. In this talk, we discuss the AI-enabled network design in vRAN, backed by several use case studies.
​
Dr Xiaoyang Wang is a Lecturer in Artificial Intelligence at the Department of Computer Science, University of Exeter. Her current research mainly focuses on machine learning, especially reinforcement learning, and its applications in next-generation network design and management.
.png)
Detecting Unexpected Behaviours in the BT 4G Network
Ed Austin
Senior Research Associate, Lancaster University
16th May 2023, from 13.00 - 13.30
In this talk we will discuss how anomaly detection is being used to identify unexpected behaviour on the 4G network. Such behaviour may be linked to events that can have a significant impact on an end user, and so it is crucial that they are identified as quickly as possible so that any appropriate actions can be taken. To address this monitoring challenge a range of different algorithms have been developed, and we shall first provide an overview of these tools before highlighting their current impact within BT.
Dr Ed Austin is currently a Senior Research Associate working on anomaly detection on the NG-CDI project. His research focuses on the detection of anomalies within streaming data, using techniques from online changepoint detection, nonparametric statistics, and functional data analysis to detect unexpected behaviour in telecommunications systems in real time. The goal of his work is to create robust methods that do not declare an excessive number of false alarms to a user, whilst also ensuring the detection is rapid enough to be of use before an issue can be noticed by a customer.
Outside of his research he am a keen runner, play guitar, and enjoy spending time with his partner and their daughter.
.
​
NetDevOps: De-risking Network Automatio
Will Fantom
PhD Student, Lancaster University
23rd May 2023, from 13.00 - 13.30
As networks become increasingly software-driven, be that as virtualised network functions or as novel configuration tooling, management practices and workflows are having to adapt. Drawing inspiration from the cloud software domain, network operators are looking to adopt streamlined autonomous workflows via a popular methodology: DevOps. This can bring about more automation and faster deployment times, whilst leveraging structured communication and specific tooling to reduce the risks associated with development to deployment workflows. That being said, networks present challenges for a methodology such as DevOps that cloud software do not, such as scale and heterogeneity. As a result, technologies and tools designed for cloud software must be altered to address these challenges. This talk will cover how tools can be developed that enable NetDevOps, specifically how CI/CD and rich telemetry systems can be leveraged in network infrastructures.
Will is a final year PhD student at Lancaster University, investigating how the DevOps pipelines found today in cloud software can be migrated and adapted to bring the benefits of automation to future network infrastructure. His work focuses on designing autonomous workflows appropriate for the reliable provisioning of network services, facilitating modern network control paradigms such as intent-based networking. His work leverages novel technologies such as digital twins and unikernels, along with more common DevOps and Sys Admin tools such as Docker and K8s.
.
​
AI-based auto-scaling of UPF instances in Containerised 5G core network slices
Abdirazak Rage
PhD student, University of Surrey
13th June 2023, from 13.00 - 13.30
Auto-scaling the use of resources is essential in optimizing service performance and reducing the energy consumed by networks. 5G transforms the traditional telecommunication architecture to Service-Based Architecture (SBA), enabling more flexible service provisioning. SBA provides a cloud-native service framework in which mobile core network functionalities (authentication, mobility management, data forwarding, etc.) are supported by Network Functions, self-contained software applications that can be run on commercial off-the-shelf hardware. SBA-based 5G core network enables deployment of the core network functions as containers on private and public cloud using Kubernetes.
In this talk, I will discuss the following topics:
-
The benefits and challenges of deploying 5G core network functions on Kubernetes.
-
5G core network slices deployment on Kubernetes.
-
AI-based auto-scaling of UPF instances in each core network slice.
-
Performance comparison of the AI-based auto-scaling method and a threshold based approach.
​
Abdirazak received his M.Eng degree in Electronics and Telecommunications Engineering from Universiti Teknologi Malaysia, Johor, Malaysia in 2014, and his B.Sc (Hons) degree in Electrical and Electronics Engineering from Omdurman Islamic University, Omdurman, Sudan in 2012. He is currently pursuing the PhD degree with the Institute for Communication Systems (ICS), University of Surrey, UK.
.png)
.png)
.png)
.png)
.png)
.png)
.png)