ETSI claims new Flexilink spec address failings of IP for advanced 5G

One of ETSI’s most important activities is working on an alternative protocol to TCP/IP, which is not fit for purpose for 5G networks, at least in scenarios like slicing, which require guaranteed reliability and responsiveness.

The Next Generation Protocols group (ISG NGP) has released its first full specification for the user plane, GS NGP 013, which promises to optimize performance, efficiency and scalability and make 5G more capable of supporting some of the key use cases envisaged for it – ultra-reliable low latency (URLLC) applications and network slicing.

Unless 5G can enable such services in an optimal way, it will lose its main differentiator, and commercial justification, compared with 4G. Yet the NPG group believes this will not be possible unless 5G comes with a new approach to packet routing.

Of course, as Andy Sutton – founding chair of the NGP ISG and a principal architect at BT/EE – explains, IP was designed for the very different hi-tech world around 1980, when memory size was the big challenge, not latency, and mobility was not an issue. Mobile IP has been a challenge in terms of performance and hand-off since cellular networks first became IP-based with HSPA, and still entails many compromises.

Hence the ETSI GS NGP 013, also titled ‘Flexilink: an efficient deterministic packet forwarding in user plane as well as packet formats and forwarding mechanisms’.

Flexilink focuses on packet formats and forwarding mechanisms for the new protocol, with the aim of achieving efficient deterministic packet forwarding in the user plane. In conventional IP networks, every packet carries all the information it needs to route it to the right destination. But this concept becomes too restrictive in software-defined networks (SDN), or those supporting control/user plane separation (CUPS). In these architectures, most packets are part of an ongoing flow such as a video stream, so there is increasing separation between two processes – deciding on the route for the packet, and actually forwarding it.

So in the new spec, there are two levels of service defined. One is ‘basic service’, which is suited to traditional statistically multiplexed packet data; the other is ‘guaranteed service’, supporting very low latency and high reliability for applications like video, Tactile Internet, or vehicle positioning.

Flexilink only specifies user plane packet formats and routing mechanisms. Specs to manage flows in the control plane have already been defined in an earlier NGP document, ‘Packet Routing Technologies’, published last year.

John Grant, chair of the NGP ISG, said in a statement: “Current IP protocols for core and access networks need to evolve and offer a much better service to mobile traffic than the current TCP/IP-based technology. Our specifications offer solutions that are compatible with both IPv4 and IPv6, providing an upgrade path to the more efficient and responsive system that is needed to support 5G.”

The ISG NGP also released two new reports to back up the claims for its protocol. One, ‘Recommendation for New Transport Technologies’, provides an analysis of TCP and other transport technologies, and their limitations; and sets out high level guidance about the architectural features required in a transport technology for 5G applications, in which the control and data planes are separated.

The report says: “With traditional transport technology, for all TCP traffic passes through a DIP router, each TCP session can only obtain a fraction of bandwidth. It is related to the total number of TCP sessions and the egress bandwidth (100 M). With new transport technology, new TCP sessions (DIP flows) could obtain the expected bandwidth or the minimum latency. And the most important thing is that the new service is not impacted by the state that router is congested, and this can prove that new service by new transport technology is guaranteed.”

The other report looks ahead to slicing, in which a virtualized, end-to-end network is created and optimized for specific services or users. The document, ‘End-to-end Network Slicing Reference Framework and Information Model’ defines a generalized architecture that would allow different operators to coordinate and concurrently operate different services as active network slices.

It aims to “provide a simple manageable and operable network through a common interface while hiding infrastructure complexities. The present document defines how several of those technologies may be used in coordination to offer description and monitoring of services in a network slice.” It describes the high-level functions and mechanisms for implementing network slicing, as well as addressing security considerations.

The new protocols and papers represent rapid progress from the publication of the group’s first white paper in 2016. Then, the group wrote: “Standards bodies such as IETF and 3GPP have been the driving force behind the success of the Internet and its integration with mobile communications; however, such organizations tend to solve problems in a segmented manner, focusing on a specific protocol layer or service requirement. NGP, on the other hand, emphasizes a holistic approach and broader scope across various aspects of the current network functions and operations.”

The NGP group aims to influence the key communications and Internet standards bodies (3GPP, ETSI, IEEE, IETF and ITU-T) to adopt its recommendations; and to drive harmonized requirements which will be optimized for multi-access communication including wireless, wired and cellular.

“Current and future use cases include 4K videos on various devices, massive IoT, drone control or virtual reality to name but a few: use cases that have nothing to do with those of the 70s,” said Sutton. “A modernized network protocols architecture had to be triggered and this is why NGP ISG was created.”

Sutton added: “The TCP/IP protocol suite has undoubtedly enabled the evolution of connected computing and many other developments since its invention during the 1970s. NGP ISG aims to gather opinions on how we can build on this momentum by evolving communication systems architectures and networking protocols to provide the scale, security, mobility and ease of deployment required for the connected society of the twenty first century.”