Transport SDN: A new flavor for software defined networking

Michael Kennedy, ACG Research

Kennedy

Software defined networking has the potential to reduce cost, accelerate service delivery, and increase the flexibility of transport networks (Layer 0 and Layer 1). The Open Networking Foundation's (ONF) initial efforts, however, have been directed at defining SDN and the OpenFlow standard for packet networks.

Having recognized that transport networks face different technical challenges and market conditions than packet networks, the ONF has established the Optical Transport Working Group (OTWG), chaired by Ciena (Nasdaq: CIEN). The optical transport group will liaise with the ONF's wireless group to avoid duplication and seek common solutions. The objective of the OTWG is to capture the benefits delivered through the automation and programmability of circuit and photonic systems--OTN and WDM equipment.

Immediate transport SDN opportunities include addressing the WAN portion of data center interconnect, multi-layer optimization of Layers 0 through 3, and managing the turn-up and turn-down of 10GE intercity services.

Data center interconnect forms a data center cluster by connecting multiple data centers over a WAN. The cluster then functions as one virtual data center. This enables reduction of each data center's peak capacity by optimizing peak capacity for the cluster--VM load is transferred from a data center that is near peak capacity to one with slack capacity. Since data center capex is driven primarily by peak capacity requirements, this reduces overall data center capex by 35 percent (my estimate). Furthermore, a Brocade (Nasdaq: BRCD) SDN WAN virtualization use case showed a 61 percent TCO reduction for the packet network. With transport SDN, further cost reductions and service velocity increases can be obtained by rapidly bringing online high bandwidth (tens of Gbps) and high quality (low latency and jitter) connections for the brief time needed to transfer virtual machines (VMs) between data centers.

Multi-layer WAN optimization simultaneously optimizes protection and capacity planning for the optical and packet layers. One of our recent reports showed that a Cisco (Nasdaq: CSCO) solution significantly reduced router and optical transponder requirements and produced a 50 percent TCO savings through integrated capacity planning as compared to separate optical and packet protection and capacity planning processes. The standards-based transport SDN approach will extend multi-layer WAN optimization to include multiple packet and optical vendor solutions and accelerate its adoption by reducing the risks associated with proprietary solutions.

Managing 10GE intercity transport services via transport SDN enables rapid turn-up and turn-down as well as rapid monetization of these high value services. The use case also will allow custom tuning of SLAs including latency and jitter. Automation and programmability will reduce the sales and provisioning cycle from weeks or months to minutes.

Transport networks face several challenges not faced by packet networks. Transport is by definition circuit-based and distance sensitive. Network design elements such as transmission line attenuation, amplifier gain, and end-to-end loss budgets must account for the condition and type of optical fiber and the color of wavelengths. Factors such as the modulation, error correction, and chromatic dispersion compensation schemes must be tuned for each path, wavelength, distance, and optical media type. Some design parameters also must be adjusted to compensate for ambient temperature and the changing physical characteristics of cables and connectors.

Transport protection and restoration processes, in addition, must be designed to function rapidly across metro and long-haul (national or global) distances. Not all transport control plane functions can be centralized. In addition, mechanisms must be established to efficiently map packets onto Optical Transport Network (OTN) ports and entities such as Optical channel Data Units (ODUs) and wavelengths.

SDN controllers directly control packet switches by populating the switches' forwarding tables via the OpenFlow protocol. Direct control of optical switches and WDM system requires an OpenFlow specification that goes well beyond path specifications and node addressing to include the transmission engineering and analog design data (discussed above). Consequently, the OTWG is pursuing two approaches: OpenFlow direct control and OpenFlow abstract control. Direct control implements a Transport OpenFlow interface and agent on each network element and talks directly to the OpenFlow Controller. Under abstract control the OpenFlow controller operates on an abstracted view of the network provided by a mediation layer. This eliminates the need for the OpenFlow Controller to have an interface to every network element. Under abstract control it is likely that network elements and/or the mediation layer will retain some vendor proprietary elements. Direct control is a fully open solution.

The OTWG has targeted April 15, 2014 for a final version of a requirements document, recommendations and extensions to the ONF protocols for optical transport networks, and a potential revision of the OTWG charter statement with all deliverables subject to review by the Technology Advisory Group Chairs.

The transport SDN initiative extends the SDN standardization program to the optical transport systems industry, one which has a $14 billion annual run rate according to ACG Research. Primary benefits are much faster and flexible delivery of optical transport services. Faster service turn-up combined with multi-layer WAN optimization also will increase utilization of both packet and optical networking assets from levels that are now well below 50 percent to the 60 percent to 70 percent range. Transport SDN is unlikely to deliver lower cost optical transport equipment, however. The highly contentious and crowded optical equipment market has driven vendor margins to low levels. Open standards are unlikely to drive equipment prices any lower.

Michael Kennedy is a regular FierceTelecom columnist and is Principal Analyst at ACG Research. He can be reached at mkennedy@acgresearch.net.