Purpose:As Broadband Networks become more dynamic with SDN control and Cloud Central office, it is now possible to programmatically control which User Plane (UP) function each individual subscriber should be connected to. This creates many advantages for an operator to offer different service propositions to different customers. At the same time, User Plane functions (such as the BNG) are becoming increasingly disaggregated and cloud native, with centralised control plane and subscriber state and the ability to scale out (add additional UP processing functions) to manage short term or long term changes in load. There is a need for a standardised approach for a disaggregated service function ,such as a disaggregated BNG (dBNG), to be able to be able to identify to which UP instance newly authenticating subscribers should be connected, or to request that existing subscribers should be redistributed or moved between UP instances. In other words, we need to define an architecture and interfaces such that the access network can offer an ingress load-balancing capability towards cloud-native user plane functions Motivation:Network Operators will not be able to effectively deploy disaggregated service functions such as the dBNG without a standardized approach to balance and move subscribers between UP instances. Service Providers increasingly desire to differentiate the services that are offered to individual customers (eg low latency / by revenue / for high throughput). This project will enable increasing differentiation by steering subscribers to a suitable UP function. This could include UP that are deployed to offer different SLA (i.e edge services). It may also include use cases where a subscriber-specific User Plane is created on demand, to which the subscriber session is then dynamically connected. Network Operators need new tools to be able to manage and upgrade networks as the industry moved to sdn/nfv. Session Steering will enable software deployment approaches in line with the cloud paradigm (such as automated incremental upgrades with canary testing on a small number of subscribers), as well as additional network resilience. Our industry is under increasing pressure to reduce power usage. The ability to dynamically move active subscribers between functions without service impact will allow hardware / software to be temporarily removed from service at certain times of the day. Scope:This project will create a WT that defines an architecture for Subscriber Session Steering, using the dBNG as an exemplary function. The following are in scope for inclusion in the project: Phase 1: - Identification & definition of the opportunities and use cases for session steering.
Phase 2: - How to identify the UP instances that can serve a subscribers requirements
- How to balance newly authenticating subscribers amongst the available UP instances that can meet their requirements.
- How to request that a subscriber or group of subscribers is moved from one UP instance to a different UP instance (without customer impact if at all possible).
- How a change in subscriber policy can trigger a change in the placement of a subscriber.
- Requirements on the SDN controller to support session steering
- Requirements for the Service Function (eg dBNG) to support session steering
- Identification of the protocols and interfaces that will be used
Note: the term 'Subscriber Session' is used within the context of this NPIF as per the definition in TR-146. It is recognized that there may be use cases for steering with a different context of session (IP session or even IP Flow), but this is currently out of scope. |