Packet-optical transport systems: Platforms for metro transformation

Packet-optical transport systems have gained rapid adoption and wide-scale deployment to support increased traffic and performance requirements for key applications globally. Scalability, performance, and economics have established these new systems as the metro building blocks of choice for wireless backhaul, business Ethernet, broadband backhaul, data center interconnect, and wholesale network initiatives. However, the architectural and operational aspects of these systems are proving far more strategic. In many cases, packet-optical transport systems can provide an ideal foundation for the transformation to software-defined metro networks.

Defining software-defined networks
Let’s start by exploring the concept of a software-defined network (SDN) and the role of packet-optical transport systems to enable a software-defined metro infrastructure. SDN has recently achieved widespread interest for data center applications. The SDN concept is simple and compelling: decouple and open the historically closed hardware and software model of conventional networking products to enable greater innovation and cost reduction.

SDNs employ a centralized control plane, with a common SDN operating system (SDN-OS). South-bound protocols, such as OpenFlow, will enable communication and management control of diverse products from multiple vendors using a common SDN-OS. A centralized control plane will enable common policies to be applied to different equipment, from multiple vendors, across the network.

The SDN-OS architecture will also support open north-bound application programming interfaces (APIs) so data center network operators and third parties can create and integrate new applications – independent of the hardware or the SDN-OS vendor. Ultimately this framework will facilitate a more open, multi-vendor, programmable network that reduces cost and fosters desperately needed innovation. While Google recently disclosed that they’ve already begun using an early implementation of SDN for their internal backbone network, SDN is still in the very early stages as an emerging architecture. Industry forums such as the Open Networking Foundation (ONF) are working to define key standards and best practices to make SDN a reality for vendors and data center network operators globally. However, it will likely be years before SDN matures enough for mainstream data centers to broadly deploy in scale. Fortunately for service providers, packet-optical transport systems enable a similar framework to realize software-defined metro networks via a phased transformation that can start immediately.Multi-Layer control plane simplifies SDN to the metro Early SDN initiatives are focused on Layer 2 (Ethernet) and Layer 3 (IP) equipment, which are prominent in data centers. A key goal of SDN is to decouple and centralize the historically embedded and distributed software control plane from the Ethernet switch and routers. This separation is required to enable hardware platforms and software systems/applications to evolve independently, and to enable multi-vendor networks with a common centralized control plane. Metro networks are similar to data centers in many ways, but more complex. Metro networks must support connectivity between end users and upstream content sources/cloud services as well as other service providers. To do this in scale, metro networks comprise:

Layer 0 (optical/DWDM)
Layer 1 (Connection Oriented Ethernet-COE, SONET/SDH, and/or G.709 OTN)
Layer 2 (Ethernet) systems, which collectively support the above as well as other services.

These additional technologies and network layers make the goal of a common, centralized control plane even more complex than in data centers.

Packet-optical transport systems have been purpose-built to support the multiple concurrent technologies used in metro networks. They support advanced Ethernet services, with COE, SONET/SDH, OTN, and/or DWDM/ROADM functions in temperature-hardened platforms sized to scale from the access aggregation edge to multi-terabit platforms optimized for handoff at the core. The ability of these platforms to perform aggregation at one layer and multiplexing and transport using lower layers of the network reduces latency and cost.

The ability of packet-optical transport systems to natively support the full spectrum of metro technologies and services in a common chassis, with optional multi-layer switch fabrics, yields material capital and operational cost savings over discrete multi-layered metro networks. However, one of the most strategic attributes of these systems is the centralized control plane that provides the common control across all associated metro-network layers. This metro-wide system capability simplifies the creation of a centralized control plane for software-defined metro networks.Model-driven architecture Packet-optical transport systems’ centralized control of multi-layer metro networks is necessary but not sufficient. Networks also must support centralized control in multi-vendor networks, and with applications that can be applied end-to-end. The network OS must also support an open API architecture to enable third parties to develop their own applications for a more programmable network. The centralized packet-optical transport system software architectures required to simultaneously support concurrent, multi-layer operations must be inherently more scalable and flexible than conventional closed-system products. Traditional development models of deeply embedded software requiring different operations constructs for different layers is too rigid and impedes multi-vendor integration. As a result, the more model-driven software architecture of packet-optical transport systems is well aligned to support multi-vendor integration and open APIs, enabling the broader software-defined metro vision. In most cases, multi-vendor integration will progress in phases. At a minimum, basic topology and alarm integration will be required to show the relationship and status among edge devices, packet-optical transport systems, and upstream core routers and/or packet-optical platforms. This level of integration will simplify fault isolation and problem resolution to improve performance. Flow-through provisioning of services is required from the customer edge to far-end handoff points, which automates and simplifies service activation across multi-vendor networks. This end-to-end, multi-layer provisioning with a centralized control plane will yield dramatic operational benefits, faster time-to-services, and improved overall customer satisfaction. Planning tools are also required to concurrently plan across multiple network layers, and across multi-vendor products that compose end-to-end service connections. Integrated multi-layer planning will yield more efficient and higher-performing network designs. With the increasing focus on service-level agreements (SLAs), multi-vendor integration is also required to monitor performance on an end-to-end basis as well as a section-by-section basis – giving service providers and their customers visibility into key performance parameters.

Finally, the model-driven software architecture of packet-optical transport systems also sets the stage for more modular and open APIs. Applications innovation may range from leveraging the system’s ability to virtualize networks (which predict key performance indicators such as latency), to enabling flexible new Ethernet services that allow for burstable bandwidth with incremental pay-per-throughput above committed information rates. The long-term beauty of an open applications model is that innovation is no longer limited by the imagination or priorities of a single vendor. While this multi-vendor, applications innovation scenario may sound far-fetched, packet-optical transport systems with many of these capabilities already exist today and are broadly deployed globally. While some of these capabilities are not yet available from many vendors, the architecture of the packet-optical transport system naturally supports the requirements and vision of a software-defined metro network.
An incremental approach to software-defined metro transformation As exciting as this new era of communications sounds, change is always difficult. The reality is no service provider can afford to simply retire all of its existing systems and immediately deploy new software-defined metro networks overnight. Instead, strategies must be established and implemented that systematically fit within short-term business constraints while positioning for the longer-term strategic vision and evolution for sustained profitability. In many cases, the short-term constraints may appear to favor a “more of the same” approach to network and capacity expansion. However, most service providers face a difficult situation in which the current cost of scaling network capacity is growing faster than revenues – leading to an unsustainable business model. Therefore, change is inevitable. The good news is that key applications -- such as wireless backhaul, business Ethernet, broadband backhaul, data center interconnect, and wholesale -- are driving major upgrades or incremental network build outs with good revenue potential. As these case-by-case network enhancements are implemented, they can provide a foundation for transformation. By establishing a clear vision of transformation to software-defined metro networks, service providers are better prepared to select and incrementally deploy point products that satisfy immediate business requirements, yet provide a foundation for software-defined metro transformation. Fortunately, innovative new packet-optical transport systems provide the rare opportunity to do both.Frank Wiener is the vice president of marketing for Cyan, a developer of purpose-built packet-optical transport systems. Prior to joining Cyan in 2008, Wiener held a range of individual contributor and executive leadership roles spanning engineering, product management, marketing, sales, and general management at Nortel, AT&T/Paradyne, and Calix.