Accelerating deployment of metropolitan optical-network services

Th 70265

Future metro-access transport platforms must open the doors for new and flexible service offerings-without all the hassle.

HAKIM DHILLA and RAMJI RAGHAVAN, Coriolis Networks Inc.

The emergence of broadband access services such as DSL, cable modem, and Ethernet, coupled with the build-out of massive bandwidth capacity in the long-haul and metro core networks, is causing the next battleground to shift to the metro-access network. Here, equipment vendors and service providers will battle for the minds and dollars of the enterprise-network community-both large and small. Who wins and who loses will be determined by who does the best job of delivering services to the user in a flexible, cost-effective, and timely manner. The economics of enterprise service delivery are being redefined and new cost points established.

Metropolitan optical transport networks today are predominantly built using SONET/SDH add/drop multiplexers. SONET/SDH is a mature and well-understood technology and has an established history of interoperability and a large deployed base. However, the poor efficiency of SONET in transporting data traffic is proving to be a major impediment in the growth of metro optical networks. A few key problems plaguing management of optical networks today are:

  • Labor intensive and lengthy segment-by-segment provisioning.
  • Manual node and/or physical port configuration.
  • Inventory management.
  • Node and network-wide upgrades.

To stay competitive, network operators must improve the ease and speed of provisioning services within and across their metro domain. They need an architecture that does for optical networks what packet switching did for private-line networks-an architecture that maximizes network use and minimizes complexity, thus improving the velocity of service provisioning and lowering operating costs.

Lowering operational costs and improving provisioning velocity will play an important role in fueling the evolution of metropolitan optical transport networks. Solutions that enable rapid deployment of new networks, easy management, simple administration, and non-service affecting upgrades of the network will be successful.Th 70265

Figure 1. It is essential that the metro control plane, unlike traditional implementations, supports control traffic routing and rerouting features in both ring as well as mesh topologies.

Provisioning services in legacy SONET transport networks can be a manual process that may require static configuration of each node and segment of the network. Fortunately, most TDM traffic patterns in legacy metro networks are a predictable minimal network design once established.

Today, however, an ever-increasing percentage of data traffic is being carried over networks. Data traffic typically requires much higher and less predictable bandwidth. That translates into big fat pipes statically over-provisioned in today's metro networks in TDM increments, such as DS-1/3 (1.55 Mbits/sec and 45 Mbits/sec, respectively) and SONET interfaces (OC-n), without regard for the bursty nature of most data traffic. That usually results in massive over-engineering of networks and consumes large amounts of precious fiber bandwidth.

Therefore, it is necessary to develop a new, innovative architecture that facilitates transport of bursty data across metro networks alongside legacy TDM traffic. Many of the lessons learned and commonly practiced in switched and routed data networks need to be adopted in next-generation metro-optical networks.

One of these lessons is using an intelligent control-plane architecture based on principles and concepts of packet-switching technology-a mechanism that treats the optical transport as a large, shared pipe for on-demand, dynamic bandwidth management instead of using fixed-bandwidth point-to-point connections. This architecture must provide a rich set of software services that introduce A-Z-Q provisioning, regardless of protocol, interface type, or topology, into the optical transport networks. It must allow unified management of all metro customer access devices and optical resources in a network-simple end-to-end provisioning of services across different network elements (NEs) from a centralized location.

A unified and centralized provisioning model in metro-access networks has several benefits. It provides a single point of control to service providers for a certain network domain, where it is simple to enforce policies. Centralized provisioning also avoids truck rolls to individual NEs for simple configuration and provisioning activities. Traditional SONET networks typically provide a "remote" provisioning capability that has similar benefits.

However, centralized provisioning also minimizes the number of management network connections needed to provision circuits in a metro network. Legacy metro-network provisioning necessitates the need for separate network connections to each NE in a metro network. This practice places an enormous burden on the element and network-management systems, not to mention the inconvenience and expense of managing several different management networks. Th 70266

Figure 2. One-step process of A-Z-Q provisioning of a metro network (a), compared with provisioning of traditional SONET/SDH metro networks (b).

Creation of metro-access domains within a network, which are managed centrally via one or more gateway network elements (GNEs), offers a scalable solution for management of networks. The equipment/ network-management station (EMS/NMS) should be able to remotely manage all the elements in one or more networks (domains) as long as the EMS has IP connectivity to the GNE. An in-band control channel eliminates the need for separate management network connections to each node in the network. It is sufficient to have a backup dial-in port and/or craft console at each NE.

Ease of provisioning is an important requirement of next-generation metro architectures, regardless of legacy TDM circuits and new packet data services. The central provisioning interface must be identical, no matter what the service type, protocol, or speed. Even the most sophisticated and involved data services should be able to turn up with a simple "point and click" operation, without requiring the technician to attend several days of school to become conversant with a new protocol.

Protocol-agnostic provisioning requires an architecture that doesn't need a two-stage provisioning process common in today's metro ring networks. Legacy SONET/SDH networks require partitioning a ring into individual STS or STM channels, which may then be individually assigned to carry TDM or data traffic. An STS carrying TDM traffic usually requires further provisioning of its virtual tributaries.

This method requires a prior partitioning of a network for TDM and data, which calls for accurate forecasting of traffic patterns-almost impossible to achieve. A metro-network architecture that allows arbitrary partitioning of the SONET/SDH ring bandwidth into TDM and bursty data-and where the partitioning may be dynamically altered as new transport circuits are provisioned-would increase the velocity of provisioning dramatically.

A control plane optimized for metro-access networks is an important component in achieving the goals of unified/central provisioning. It must be de signed for easy turn-up of new networks, simple installation of new NEs within an existing network, and rapid provisioning of new services.

This metro-network control plane should be optimized to operate in ring-based topologies of today's intra- and inter-metro optical networks, but must also scale to enable end-to-end provisioning across mesh networks. The following elements constitute the functionality of the control plane optimized for metro-access networks:

  • Signaling protocol for end-to-end connection setup within a metro-access ring/domain.
  • Standards-based signaling protocol (Generalized MPLS) for connection setup across different metro-access domains.
  • Routing protocol.
  • Neighbor discovery and link management.
  • Network topology discovery and state distribution.
  • Policy-based path selection.
  • Support for traffic engineering.
  • IP-based operations, administration, maintenance, and provisioning access to every NE in the metro network.

An aggregate of all these attributes is needed to achieve the goals of substantially in creasing the velocity of service provisioning and significantly reducing the operational costs of deploying, managing, and scaling a metro-optical network.

The first step in simplifying the operation of turning up of new networks and adding new NEs to an existing network is to have an automatic topology discovery mechanism in the control plane. A ring-optimized routing protocol is necessary for automatic discovery and configuration of new nodes installed in a metro network as well as connection routing of new circuits in the network. Automatic rerouting of control messages around ring failures must establish a resilient control path for all software applications. The routing protocol is used for the following functions:

  • Routing of signaling messages between NEs.
  • Routing of management traffic (transaction language 1-TL-1 and simple network-management protocol-SNMP) from the GNE across the ring to different NEs, enabling a single management (configuration/monitoring) point for an entire network through the GNE.
  • Software upgrades of any or all nodes in a metro network via the GNE.
  • Path selection-an intelligent constraint/policy-based path selection algorithm that may be used for traffic engineering within a metro network.

A routing manager automatically discovers new nodes that are plugged into an existing ring without requiring elaborate configuration of the new node. The routing manager simplifies the installation of new nodes:

  • Ring and NE discovery.
  • Node identification assignment.
  • IP address assignment to node interfaces.
  • NE plug and play capabilities.
  • Loss of connectivity sensing.
  • Location of improperly connected ring ports.

Service providers should be able to plug a new NE into an existing network with minimal manual configuration required. The rest of the provisioning and configuration should be driven from an easy-to-use graphical user interface (GUI)-based EMS through the GNE.

An offline configuration mode, or design node, should be available to allow a node and associated transport channels to be completely provisioned in the GNE before turning up the new node in an existing network. This feature could significantly reduce the time required for live network upgrades.

The routing protocol must ensure that the control path necessary for automatic provisioning and state distribution is always available as long as the network has multiple paths between two nodes. The control path is used for automatically routing management traffic (TL-1, SNMP, etc.) from the EMS/NMS interface via in-band mechanisms to other NEs in the ring. The IP-based control path is also used for network-wide software upgrades from a central point without disruption to traffic.

The control traffic is automatically rerouted in the event of fiber cuts or node failures. As long as the network has a multiring configuration and a path is available in either direction from one node to another, the control path may be retained. It is, therefore, essential that the metro control plane, unlike traditional implementations, supports control traffic routing and rerouting features in both ring as well as mesh topologies. Figure 1 shows the steps required for IP-based control path.

The automatic control path establishment must provide an auto detection feature that makes it possible for the GNE to detect a bad ring configuration. That significantly simplifies installation of new nodes in an existing network or deployment of new networks.

Metro-network transport channel provisioning must be a simple, single-step process, regardless of whether the transport channel (TC) is a TDM circuit or a "burstable" data private-line service. Provisioning via the GUI or TL-1 command-line interface (CLI) should require only that two endpoint access ports, the traffic descriptor, optional traffic engineering, and class of service attributes be specified. Figure 2 illustrates the one-step process of A-Z-Q provisioning of a metro network, which the new control-plane mechanisms must enable, and points out the differences between that and provisioning of traditional SONET/SDH metro networks.

Transport connections and their associated service-level agreements (SLAs) are defined by a set of attributes that include endpoint addressing, traffic descriptor (committed rate, burst rate, burst size), class-of-service (CoS), and traffic engineering attributes. These attributes must be communicated to the associated NEs involved in the transport of the channel during connection setup via signaling messages.

Management of metro transport channels must be as rich as those of virtual circuits or label-switched paths in frame relay, ATM, and MPLS networks. The control plane signaling and call management software entities should support the following basic features common in data networks:

  • Connection admission control.
  • Path selection based on contract and constraints/policies.
  • Signaling messages for setup and teardown of connections between any two NEs.
  • Connection management.
  • Operations-administration-maintenance (OAM).
  • Connection resiliency-the ability to reroute connections around failures by preempting lower priority transport channels.

The control plane should support local management protocols and connection management protocols at the various endpoints of transport channels or circuits. Frame relay defines link-management protocols for both the user-to-network-interface (UNI) and network-to-network interface (NNI). Any changes resulting from the protocol are reliably conveyed to the opposite endpoint of the TC as required by the protocol.

Similarly, for ATM, the generation or suppression of alarm indication signal (AIS) is supported as well as frame relay/ATM interworking TCs as defined in Frame Relay Forum-8 (FRF-8). Link-management and connection-management mechanisms allow rapid detection and reaction to failures, in addition to serving as a performance-monitoring and connectivity-checking tool. In the absence of such protocols, failure detection and reroute are limited to higher-level protocols like IP, which may take seconds to recover.

Upgrading a system with new software/firmware is often a lengthy and time-consuming process. Metro control planes must enable automatic distribution of new firmware to NEs across a domain, with no impact on the data flowing through them. The firmware upgrade initiation should be a simple point and click from the EMS GUI or the TL-1 CLI. The domain-wide upgrade must be possible, even if the GNE is the only NE that has external IP connectivity to the administrative file transfer protocol server.

This configuration eliminates the need for NE connections to an Ethernet maintenance network infrastructure and cuts the upgrade procedures to a single step. Service providers typically have maintenance windows, in which they would like to upgrade a certain portion of their network and have it up and running in as short a time as possible. However, the upgrade procedures and protocols must also be flexible enough to allow selective portions of a network, or even single NEs, to be upgraded, if necessary, with the same ease.

Next-generation metro network control-plane architectures must allow maintenance of existing elements in the network with minimal disruption. Traffic engineering hooks and controls that a service provider may use to control the flow of traffic via different portions of a metro network (ring or mesh) allow a graceful mechanism to offload and isolate certain portions of the network at opportune times.

Maintenance and upgrade of existing optical networks may require isolation of a particular node from a wavelength or ring, isolation of a particular ring, or perhaps even isolation of an entire NE.

These controlled mechanisms prevent the necessity of inducing a ring failure or ring cut, thereby triggering a massive protection switch. Planned maintenance and servicing of NEs should have minimal effect on network performance.

The balance-sheet metric for measuring carrier performance is return on capital invested. Bandwidth is the most expensive asset deployed by carriers. However, this asset is being significantly under-used because of bifurcated applications across voice and data. This lack of integration not only results in higher deployment costs, but also higher costs to maintain and operate multiple networks.

The metro-optical transport services are rapidly migrating from the low-speed, static private lines of today to the high-speed burstable private lines, fat pipes, and tiered services of tomorrow. The metro-access transport platforms of the future must open the doors for new and flexible service offerings, without requiring expensive network over-engineering and complex network provisioning.

An innovative optical-network architecture based on principles and concepts of packet-switching technology provides the best solution to improve bandwidth use over an integrated network, improve velocity of service provisioning, and lower operating costs.


Hakim Dhilla is co-founder and vice president of product marketing and Ramji Raghavan is co-founder and chief technology architect at Coriolis Networks Inc. (Boxborough, MA). They can be reached via the company's Website, www.coriolisnet.com.

More in Packet Transport