Telecom networks have long been dedicated networks for Fixed, mobile or broadband communications with even some special implementations for government, policy or security departments. Lately, with IoT, some new technologies around Low Power WAN(LPWAN) came out leading again to some dedicated networks such as LORA or Sigfox.
With 5G, the target is clearly to get rid of these dedicated networks and move to a flattened network with a set of resources: access, transport or core, allocated on demandto serve different use cases, and deployed either close to the end user device, at the edge of the network, or more in the backend data center.
Fig 1- Software Defined Network towards a flattened network
NFV and SDN, 1st step towards softwerization of the telco network
A few years back Network Function Virtualization (NFV) and Software Defined Network (SDN) were introduced in the Telecom environment by ETSI and ONF respectively.
NFV was mainly focused on decoupling Hardware from Software and evolving monolithic network functions towards cloud-based software functions, centrally management by a Management and Orchestration (MANO) layer that would allow dynamic provisioning and scaling.Virtualized Network Functions (VNF) would interconnect via Virtual Switch that could be controlled by SDN controllers. Similarly Point of Presence (PoP) of virtualized infrastructure would also interact via SDN WAN (Wide Area Network) as explained in ETSI NFV EVE005. As an example below, vIMS and vEPC could be in PoP-1 and vRAN in PoP-2, all virtualized and connected via SDN.
The ETSI NFV reference architecture and standard APIs apply for fixed and mobile networks, including access and core NFV and SDN are now the baseline for 5G architecture.
As for SDN, dealing with the decoupling of User Plane and Control Planeand programmatic access to forward packets dynamically, it’s been adopted by 3GPP in release 14 for the mobile Core to separate the SGW and PGW user plane and control plane and introduce a new protocol between the two, not openflow, but an evolution of GTP-C called PFCP (Packet Forwarding Control Plane). SDN and the decoupling of User Plane and Control Plane is now the baseline for 3GPP 5G Service Based Architecture (SBA).
Fig 2. 5G SBA leveraging NFV and SDN
Softwarization of the RAN
While some of the most active NFV projects have been around LTE-4G vEPC and virtual mobile core evolution towards 5G SBA, the RAN is alsoevolving with Cloud RAN, virtual RAN, open RAN projects and 3GPP 5G RAN architecture. The focus is to evolve the RAN into a more open, software-based and disaggregated platform with the antenna as a physical element, and the rest software based that can be deployed co-located with the antenna or distributed between the edge and the core.
The 5G RAN is implementing this decomposition of the RAN into RU (Radio Unit), DU (Distributed Unit) and CU (Central Unit), each composed of a set of network function such as MAC layer, scheduler, user plane or control plane protocol layers.
Fig 3. 5G RAN decomposition in an E2E 5G Architecture
Heterogeneous Access and Slicing to best serve Verticals
With 5G, the 5G Core not only serves the RAN, Radio Access Network for mobile-cellular networks, but it also serves other access types: Wifi or fixed and broadband access. 5G also embraces unlicensed spectrum for IoT or private enterprise networks, like CBRS. The objective is really to channel all the traffic to a common core and avoid parallel independent silo networks. There are cost benefits to this approach, capex and opex, but also benefits in terms of customer experience: the same device connecting to an application could move from a cellular network to wifi or to some the unlicensed spectrum while keeping the connection to the network and service alive. This implies some multi-band dual connectivity capabilities on the device and proper implementations on the back end of course.
While softwarization and programmatic access bring flexibility, facilitates connectivity and mobility across this overall network, Quality of Service (QoS) is the most important differentiator for telecom operators to satisfy customer experience and monetize their capabilities. In order to ensure this QoS and QoE (Quality of Experience) not only for mobile users voice services or more advanced gaming users but also for enterprise and government highly demanding services in terms of bandwidth, latency or privacy, 3GPP has introduced the concept of network slicing.
Fig 4. 5G E2E Network Slicing to best serve industry verticals
A network slice is an E2E allocation of network resources between a set of devices and some applications to offer a virtual private network to different tenants.
In the example on Fig 4. We have 2 slices, one for Fixed Wireless Access (FWA) and one for Connected Cars. Some of the RAN resources are shared, ie Radio resources, while others are dedicated, ie CU PDCP-C for instance. Similarly on the mobile core, some virtual network functions would be shared such as the NSSF (Network Slice Selection Function) or the NRF (Network Repository Function) while others will be dedicated to each of the slice: typically they could have a target AMF (Access Mobility Function) and SMF (Session Management Function) to control traffic via some dedicated UPF (User Plane Functions). This way FWA traffic is isolated from Connected Car traffic.
The way 5G network have been specified with decomposition of the network functions, virtualization and SDN, allow many different compositions and flexibility in terms of deployment and business models. This is one of the most interesting aspects of 5G! Of course management and automation will be key to make this work!