Event Report: CommNet2 & Middlesex University Future Networks Workshop

Event Report


On January 16th 2017, CommNet2 and Middlesex University hosted a successful workshop focusing on Future Networks. Here, our Event Champion Dr. Glenford Mapp of Middlesex University reports on the day’s presentations and discussions:

Glenford

Placeholder

Presentations

Giles Heron (CISCO)

GilesGiles looked at the mechanisms and techniques needed to build and manage future networks using Software Defined Networking. His main point was that the promise of SDN was to reduce the cost of managing networks and not about specific mechanisms such as OpenFlow. But this must be done in a way that allows network management applications to be built and tested. He concentrated on two main tools that are being developed and supported: YANG, which is a data model used to express how data is represented and NETCONF, which is a mechanism that defines the relationship between devices and services. So one can think of YANG as a MIB while NETCONF, like SNMP, is used to implement services using RPCs with YANG as the data model. It was a very useful presentation as it tried to get beyond the SDN hype.

Download Giles’ presentation: Future Networks 2016 – Giles Heron

Professor Jon Crowcroft (University of Cambridge)

JONJon first looked at the success of the Internet, which he put down to a large slice of luck; therefore projects that concentrate on trying to basically replicate the functionality of the Internet but trying to do it better rarely work. He also said that top-down approaches rarely succeeded. Instead of trying to make wholesale changes, he said that it might be better to look at specific problems that needed to be solved. He posed these problems as challenges that were based around needed functionality to support new features such as mobility or new technologies such as IoT. Jon then looked at the interaction or the lack of between the Electronic Engineering (EE) and the Computer Science (CS) communities. He felt that effort was needed to get the communities talking and working together on real problems. He suggested that the communities do joint bids on problems to which both groups needed solutions. He said a key area that he felt should engage both groups is the area of networked machine learning and its possible application to large-scale network problems.

Download Jon’s presentation: Future Networks 2016 – Jon Crowcroft

Stuart Revell (5GIC)

sTUART
Stuart’s presentation was centred on an Interim Report from Future Communications Challenges Group, which was charged by the government to come up with a UK strategy and plan for 5G & Digitisation. The report proposed a new organisational structure based around SME & Micro Engagement, Knowledge exchange & Dissemination, a Standards Groups, which is split into key industries/sectors. A new subsystem was added based around funded testbeds and trials, which would feed into the Standards Group. The testbeds would be run using a hub-and-spokes model looking at different types of networks and technologies. The report identified 4 key areas for networking: High Speed Broadband (HSB); Network Availability, Capacity and Coverage (AC & C), Massive Internet of Things (MIoT), Critical Internet of Things (CIoT). The context of low power as well as seamless public and private engagement were deemed to be essential. He divided the 5G effort into 3 areas: 5G Radio, 5G Core and 5G Fabric. He then showed the huge positive economic impact that would be obtained by a successful implementation of the overall system. He then looked at SDN, slicing and the 5G Fabric, which acts like a Network of Networks, to provide the communications needed in key areas. Stuart felt that testbeds and trials were needed now.

Download Stuart’s presentation: Future Networks 2016 – Stuart Revell

Dr. Marco Ruffini (Trinity College Dublin)

MARCOMarco Ruffini looked at optical networks in the 5G era. He acknowledged that 5G had to be more than an incremental improvement on 4G and that optical networking could make 5G excel in the areas of enhanced Mobile Broadband and ultra Reliable and Low-Latency Communications. Fibre makes Cloud RAN possible and leads to an effective splitting of the LTE stack resulting in a radical increase in fronthaul speeds. This in turn will lead to network convergence which Marco argues should yield reduced overall network costs as well as increased performance. For Fixed mobile convergence, Passive Optical Networks (PON) should lead to a more flat overall structure. Marco then highlighted the need to look at services on the back of all the network improvements, which should lead to a more service driven way of operation. The clear separation of services and architectures should enable some novel solutions to problems, such as the old Telco Central Office re‐architected as a Data Centre using network virtualisation.

Download’s Marco’s presentation: Future Networks 2016 – Marco Ruffini

 


Discussions

After these talks there was a session in which the attendees asked technical questions of the speakers based on their presentations. These talks were really wide-ranging and hit on several issues and approaches: this diversity stimulated a lot of discussion. This was followed by lunch and a panel discussion, which discussed ways of getting better collaboration between CS & EE groups. Some points from that discussion:

  • Though the bottom-up approach is good, there is also a need to have a vision that pulls projects out of the laboratory, into real testbeds and then into commercial products and services.
  • There was an acknowledgement that the UK is good at basic research but we are less good at making our research into products and successful companies.
  • There was a discussion on whether there needs to be a search for killer applications, which would be used to gain acceptance and commercial success. The response was that it was very difficult to choose killer apps and it would be better to provide flexible and dynamic platforms and networks that would allow different types of applications to be built.
  • Another discussion revolved around the need for Cloud platforms and support for easy deployment of services.
  • There was also a recognition that it cannot just be about academic/industry: government and organisations such as TfL, Highways UK and Network Rail, also need to be part of what is happening

Recommendations for EPSRC

Attendees were then asked to generate a list of suggestions to EPSRC on two large whiteboards.

  • More meeting and exchanges between the groups. These meetings should be a real mix-up of EE & CS, industry, non-profit and government.
  • Showcase/Demo Days. These should be days where people can show the community what ideas they are working on and get feedback from the community. This could involve several departments from one or two universities that are in the same region.
  • Join the Dots: Help get local councils and government involved to build bigger testbeds: Bristol is OPEN is a good example allowing researchers to move their testbeds out of the lab and into the community.
  • Sponsor a number of large testbeds on different things: 5G, Smart Cities, IoT. Use these testbeds to create large data sets. Analyse the data sets using the HPC Centre and feed these results into standards.
  • Connect testbeds together so that we have a national testing platform.
  • Using this national platform to look at high-level issues such as using WAN information to do better network management.
  • Finally, look at machine learning (ML) and then AI for building better networks.

Leave a Reply

Your email address will not be published. Required fields are marked *