Segment Routing Circle

Segment Routing in 5G transport

  • 1.  Segment Routing in 5G transport

    Posted 11-01-2020 01:21

    To fulfill 5G promises like increased capacity and ultra-low latency, new techniques are being introduced in the 5G radio in terms of modulation schemes, encoding, channel size increase form 10-20 MHz in 4G to 100-400 MHz in 5G, channel aggregation, high degree of parallel transmission (e.g. MIMO  multiple input, multiple output antenna systems, going up to 64 MIMO), etc. This forum is about Segment Routing in 5G networks, so we are not going to discuss all the radio optimization details here.

     

    Apart from radio optimization, another toolset to address increasing capacity and strict latency demands is opening of new frequency bands (i.e., Frequency Range 2 - FR2, called often as well millimeter Wave - mmWave, with 50-400 MHz channels sitting in 24-52 GHz frequency ranges) for 5G deployments.

     

    Why I am mentioning all the above? As it has strict implication on the transport. FR2 (mmWave) frequencies, while providing huge increase in the channel capacity (10-20 MHz  50-400 Mhz, so radio channel capacity increases by the factor of 5-20), have very low reach. We are talking here about few hundred meters (100-200 meters). And, this fact has very big impact to the 5G transport network design, as the number of endpoints (radio units) that must be interconnected will increase dramatically.

     

    Therefore, 5G transport network will contain a large number of routers (CSR – Cell Site Routers), a mix of WAN and data centers components, with requirements to support traffic engineering, therefore the underlay control plane will be typically a collection of independent routing domains that interact in a collaborate fashion, as outlined in the following figure.

     

     

    5g-transport.png

     

    This, in turn, implies distributed MPLS control plane is used to spread information required to deliver the packets through the 5G transport network. It typically includes following information:

     

    •           Topological information (nodes and links in transport network, including attributes, like metrics, IP prefixes administrative groups, SRLGs, etc., associated with these topological elements)
    •           IP prefixes
    •           MPLS transport label information

     

    Based on the collected information, control plane responsibility is to

     

    •           calculate paths (shortest path to the destination, or path fulfilling certain administrative constraints, which are not necessarily the shortest paths – for example low latency path)
    •           program data plane with information required to forward the MPLS packets through the network (next-hops, MPLS labels stack manipulation rules – pop, push, swap)
    •           calculate and program in the data plane backup paths required for rapid protection mechanisms (fast re-route – FRR)

     

     

    Generalized 5G transport infrastructure can be summarized as follows:

     

    1.          Each routing domain has an IGP for internal connectivity.

     

    1.          Each routing domain has a protocol for distributing label information. Typically, SR extension to an IGP, but for some brown-field deployments, interaction with legacy protocols, like RSVP or LDP, should be provided.

     

    1.          Mechanisms to calculate Traffic Engineered paths within a routing domain. Depending on scale, TE paths can be calculated on the transport devices themselves (distributed constrained shortest path first – D-CSPF – calculation, with lightweight TE implemented via SR Flex-Algo, and more ‘heavy’ TE implemented via traditional D-CSPF), calculation can be off-loaded from less powerful cell site routers to more powerful transport devices (on-box PCE), or even dedicated servers could be deployed (off-box PCE). The diagram shows an example of distributed “Path Computation Element” (PCE) layer with “Path Computation Element Protocol” PCEP running between the cell site routers and the PCE

     

    1.          Mechanism to establish end-to-end MPLS LSPs between IGP domains. To achieve highly scalable architecture, few mechanisms could be considered:

     

    • Not requiring PCE
      • Seamless SR Architecture with BGP Classful Transport (BGP-CT)
      • Express Segments architecture with D-CSPF on CSR
    • Requiring PCE
      • Express Segments architecture with distributed PCE layer, implementing PCE function on pre-aggregation routers or on dedicated devices

     

    1.          Mechanism to convey topology and network state information from the network to the PCE and other management elements. BGP Link State (BGP-LS) and Telemetry feeds are recommended tools to fulfil this requirement.

     

     

    This post is to open the discussion regarding SR in 5G transport networks.