Using SDN to simplify G.fast deployments
Innovations in copper access networks like G.fast micro-nodes and Reverse Power Feeding (RPF) are coinciding with the wider industry move towards Software Defined Networking (SDN). This article explores how SDN enables solutions such as Zero Touch Provisioning and offline management to greatly simplify G.fast deployments. The resulting agility and efficiency enables operators to deploy G.fast wherever they need it and eliminate complexity while managing large numbers of micro-nodes. Nokia is working with the Broadband Forum (BBF) on defining standard models for access products based on NETCONF/YANG. The standardization will result in a SDN enabled copper access network that makes deploying G.fast ultra-broadband a reality for operators.
As demand for ultra-broadband continues to accelerate, service providers are calling upon the last drop of their copper infrastructure to provide ever-greater levels of performance. Thanks to digital subscriber line (xDSL) technologies like VDSL2 vectoring and G.fast, bitrates of 100 Mb/s up to 1 Gb/s are now within reach. G.fast in particular promises fiber-like speeds and is being deployed by operators as part of a deep fiber strategy that brings fiber closer to customers. Smaller and more discreet access nodes, known as distribution point units (DPUs), are then used to connect customers using the last few hundred meters of copper access in deployments known as fiber to the distribution point (FTTdp) or fiber to the building (FTTB).
There are two characteristics of a G.fast deployment that pose new challenges for network operators. Firstly, the number of nodes to be managed is significantly higher than for a traditional copper access network. Secondly, these nodes often need to be placed in non-traditional and hard-to-reach locations: in the basements of buildings, on telephone poles, under manhole covers, in pedestals or attached to external walls of homes. Many of these locations do not have a power supply. To get around this, reverse power feeding (RPF) from the customer premises is used, but this means that nodes may not always be powered on and, hence, available to the network.
EVOLUTION OF COPPER ACCESS NETWORK
Deploying FTTDp and RPF presents some new challenges for operators.
- Scale. A traditional DSLAM typically serves between 100 to 2,000 subscribers whereas a DPU will serve 4 to 50. This means the network has to scale to manage up to 500 times more nodes compared to a traditional copper access network. In order to deploy, configure and operate a network with a large number of nodes that are not always easily accessible, the nodes need to support Zero Touch Operation where post-installation visits are not required to add new subscribers or to enable RPF.
- Agility. Similarly, the large number of nodes in combination with a large variety of configurations mean nodes should support Zero Touch Install to accelerate deployment and reduce the time and expertise required of field staff. Additionally, the node and network should dynamically adapt to any network problems e.g. a switch-over due to a fiber cut, aggregation node failure, and so on. Nodes need to support auto-detection and reestablish connectivity with the management system without manual intervention.
- Abstraction. Achieving agility in deploying services across a large number of nodes requires a programmable interface to the network to effectively activate, provision, monitor and troubleshoot the underlying nodes. Standardized and open interfaces that provide abstraction and separate the management layer from the underlying hardware and provide a uniform interface to the OSS/BSS are a key requirement.
- Persistent store and cache. Traditional DSLAMs were designed for installation in the central office or in street cabinets that have access to power and are “always-on”. DPUs using RPF do not have the same luxury and may be offline. Even when offline, operators still need to be able to rollout service configurations and perform pre-provisioning of lines. This requires a persistent store and cache to reside outside the DPU so the DPU is “always-on” for provisioning even if unpowered and syncs with the updated configuration when it is powered on.
Software Defined Networking enables more agile, manageable, dynamic and cost-effective networks. As such, it can support the new dynamics required by G.fast deployments. The primary tenets of an SDN architecture are:
- Programmability. Networks are controlled by software functionality, allowing network operations to be automated and adapted in a flexible way.
- Plane separation. Separation of the management plane from the forwarding plane allows new services and behaviors to be introduced across underlying hardware.
- Abstraction. Network devices are abstracted from the control layer to ensure portability and future-proofing of investment in services, with the network software residing in the control layer.
- Central control. Centralized network intelligence allows decisions to be made based on a global view of the network, allowing rapid network changes and rollout of network services.
- Open standards. Open standards and open APIs for programming the network enable innovation and differentiation by operators.
SDN ENABLEMENT OF COPPER ACCESS NETWORKS WITH NETCONF/YANG
Nokia is working with the BBF on an SDN-enabled copper access network standard based on NETCONF and YANG to address the evolution of copper access networks.
NETCONF, developed by the Internet Engineering Task Force (IETF) as RFC 6241, is a network management protocol optimized for nodal integration into management systems. NETCONF can co-exist with CLI, SNMP, SYSLOG and other current management protocols and provides an extensible mechanism based on XML through which network devices can be managed, configurations can be manipulated and operational data can be retrieved. The protocol contains four conceptual layers.
- The secure transport layer provides built-in security for the communication path between a client (management system) and a server (network element) e.g. Secure Shell (SSH) and Transport Layer Security (TLS). This includes authentication, data integrity, confidentiality and replay protection.
- The messages layer provides a transport independent framing mechanism for configuration data, state attributes and notifications. This allows the device to expose a full formal Application Programmable Interface (API) towards the management system.
- The operations layer defines the base protocol operations to be invoked with XML encoded parameters e.g.
– retrieve device state information, edit-config – create, delete, merge or replace contents in a configuration store. This includes support for versioning, discovery of extended device functions and capabilities at runtime and transactional semantics allowing bulk operations and rollback across multiple devices in network.
- The content layer contains models for configuration and state which are realized in YANG.
YANG is a data modelling language for the NETCONF protocol that supports the hierarchical organization of data structured as modules and submodules. Reusability is achieved by allowing modules to import data from external modules either conditionally or completely or by extending (augmenting) existing modules. YANG models describe the constraints to be enforced on the data restricting the values of the nodes.
Figure 1 NETCONF/YANG Leading the way to programmable access networks
In NETCONF, the content layer is the only layer that is not standardized. However, the combination of NETCONF, YANG and standardized open data models can address the challenges of scale, agility, abstraction, and persistent store and cache in a G.fast deployment. The BBF plays an essential role in defining open architectures based on NETCONF and YANG and standardizing these network evolutions in TR-301 - Architecture and Requirements for Fiber to the Distribution Point.
Figure 2 BBF DPU Management Architecture
TR-301 introduces the concept of the persistent management agent (PMA) to address the challenge of persistent store and cache. A PMA acts as a management proxy allowing the OSS to perform operations on a DPU regardless of its availability: PMA enables an “always-on” presence of the DPU in the network. The interface between the PMA and DPU uses NETCONF with open standard YANG models being defined by the BBF. In terms of SDN, PMA provides disaggregation of the management layer and abstraction of the DPU via standard YANG models.
The PMA also addresses scale by aggregating several PMA functions into a single aggregation function (PMAA), providing a single interface towards the OSS for a large number of DPU in the network. This improves performance for bulk operations in the network and at the same time enforces best practices in operations such as data separation, versioning and abstraction in a multi-vendor environment.
The challenges of abstraction and agility require defining standard YANG models to act as a common programmable API for the network. This API can then be used to adapt the network to changing underlying conditions.
For any innovation in networking technology to succeed for the good of consumers and operators, it’s important for industry bodies like the BBF to set standards that the industry can agree and adopt. Nokia is actively working with the BBF to define the standard models required for DPU operations including G.fast, Forwarding, VLAN, QoS, Multicast, Bulk Data Collection and Software Image Management, among others, and to bring to market products based on NETCONF/YANG for SDN enablement of next generation copper access networks.