technology & profit
Crossing the electro-optical boundary
A coalition of more than 100 vendors and service providers is working to establish protocol and interface specifications that will allow IP routers, ATM switches, and other equipment to request bandwidth on demand from the optical domain.
BY AMY COPLEY
Market forces continue to drive the need for more scalable and agile network infrastructures. A major step toward this goal is achieving interoperability at the boundary between the electrical layer and optical layer of next-generation networks. That would allow service providers to support bandwidth on demand and other enhanced high-speed services.
The emergence of intelligent optical switches is heightening the focus on the issues surrounding interoperability at the electro-optical boundary. Optical switches provide electrical devices at the edge of the network with enhanced functionality-enabling point-and-click provisioning-and a mesh-based architecture at the optical layer. Along with advances in Internet Protocol (IP)-based traffic engineering and constraint-based routing protocols, which enable devices such as IP routers and ATM switches to determine when and where they need additional bandwidth, intelligent optical switching is introducing new models for dynamic provisioning of high-speed capacity.
Vendors and carriers in the industry have also started to take action. A standards initiative, the Optical Domain Service Interconnect (ODSI), was spearheaded by Sycamore Networks in January to serve as a catalyst to define the hand-off of information between the IP and optical layers at the core of the public network. To keep pace with market demands, the ODSI coalition came together to collaborate on a practical technical framework based on extensions to existing protocols and interfaces.
Formation of ODSI
While the goal of interoperability between the two layers is widely supported, there are two fundamental architectural approaches to binding the optical and electrical domains; each has its advantages and disadvantages. In the overlay or user-network-interface (UNI) model, the user device (IP router, ATM or frame relay switch) communicates with the network device (intelligent optical crossconnect, intelligent optical transmission equipment) using a well-defined interface. This approach allows the two domains to communicate but does not require a complete topological exchange between devices. The peer model assumes a common signaling protocol running across the electrical and optical layer. In this model, the user devices maintain the topology and control the optical network.The ODSI initiative kicked off the industry effort to define an open UNI-type mechanism that will enable electrical devices such as IP routers, ATM switches, and SONET/SDH add/drop multiplexers to request bandwidth on demand from the optically switched network. That would allow high-speed capacity to be provisioned in an agile and dynamic fashion. Subsequently, other industry groups such as the Optical Internetworking Forum (OIF) and the Internet Engineering Task Force have also begun evaluating and proposing recommendations for both the UNI and peer approaches.
 |
| Figure 1. A surge in unpredictable traffic from Internet users requires new data-centric network architectures. |
|---|
While other standards bodies have begun organizing workgroups to address this issue, the ODSI was the first and is making rapid progress toward completing its functional specification by year's end. The ultimate goal is to make meaningful contributions to the development of standards in the form of a draft recommendation and published results from interoperable solutions testing.
Since its inception in January, ODSI's progress has been nothing short of phenomenal. The coalition, comprising more than 100 network vendors and service providers, has made tremendous progress on the functional specification and accompanying protocol documents, including the management information base (MIB), signaling specification, common open-policy service (COPS), and service discovery documents. After the third ODSI meeting July 25 in San Francisco, the coalition is about 85% complete with the functional specification protocol document. Updates have been provided to the OIF after each meeting. Documents can be downloaded from www.odsi-coali tion.com.
This accelerated pace is a promising sign for architects of greenfield networks as well as network planners of legacy infrastructures. However, the best next-generation architecture will depend on a number of factors-no one-sized solution fits all. Given the simplicity of the UNI model and progress being made in the protocol specifications, this approach is likely to be first-to-market, enabling some of the functionality service providers need within their heterogeneous networks.
Most agree the optical domain is migrating from today's ring-based transmission infrastructures to more scalable, flexible mesh-based network schemes. Advances in hard optics such as Raman amplifiers, tunable lasers, and optical interconnect technologies continue to fuel innovation in the transport, switching, and management systems that form the foundation of optical networking.
 |
| Figure 2. Next-generation architectures must support the enormous traffic developing from the increasing use of higher-speed access technologies. |
|---|
At the same time, the higher-layer electrical service domain, with its core terabit routers and ATM switches, is changing so service providers can adapt to the growing volume of IP traffic flows and data service requirements. As the number of Internet users grows dramatically worldwide, the volumes of unpredictable traffic in the backbone hasten the need to design data-centric network architectures (see Figure 1). These changes must support the enormous amounts of traffic developing from the use of higher-speed access technologies such as xDSL and cable modems (see Figure 2). The boundary between the electrical and optical domains is at the center of this change and the focus of the ODSI.
Bandwidth on demand
On-demand creation of network bandwidth is not a completely new concept; it is possible today through ISDN, ATM, and frame relay switched virtual circuits (SVCs). The main difference is that ODSI is focused on larger bandwidth channels (e.g., OC-3, OC-12, OC-48, OC-192, Gigabit Ethernet) and will be used by devices that are already aggregating heavy traffic.Since bandwidth-on-demand schemes have been around for many years, there are a number of protocols that can be used to signal bandwidth requests to the optical network. However, these existing protocols must be modified and enhanced to support the optical network's unique characteristics. Since TCP/IP is the universal control protocol for data communications, using IP addressing and protocols is a natural choice. These extensions form a key element in the ODSI framework.
In the ODSI functional specification, the service is the dynamic creation of point-to-point bandwidth channels between user devices. There are three components to the service definition:
- Specifying the end-points of the bandwidth channel.
- Determining the characteristics of the bandwidth.
- Determining what actions can be requested on the bandwidth channel.
The first phase of ODSI interoperability testing is scheduled for the November time frame. Vendors with ODSI-compliant code in a switch, client, or third-party product can participate in the testing. The next full ODSI meeting, which is expected to wrap up the informal coalition, is scheduled for January 2001.
Completed ODSI specification and implementation
When the ODSI specification is completed, vendors will be able to download the code from the www.odsi-coalition.com Website for free. The finished specification will provide vendors with the tools to be complaint with ODSI real-time, high-speed bandwidth and automation. When ODSI is incorporated into vendors' equipment, the complaint devices can work together to allow carriers to offer more flexible services in real time.The recent Summer Olympics in Australia is an example of a situation where bandwidth demand increases substantially for a short period of time. For instance, the network televising the event in the United States could have leased a number of OC-48/STM-16 circuits from a carrier in Sydney for the duration of the event. The television network also could have leased an amount of capacity in the United States in major metropolitan cities. If the U.S. women's soccer match were Webcast and the major sites were congested, the television network could have dynamically increased the bandwidth as necessary during the event by having additional server sites available in smaller regional sites to offload the major cities. When the soccer match ended, the additional bandwidth could have been torn down.
In that scenario, without the availability of point-and-click provisioning and dynamic bandwidth allocation, a customer would most likely have to sign a contract for 24 months, and it may take three months prior to the event to provision the additional bandwidth. The dynamic creation and teardown of bandwidth allows a more flexible and cost-effective option for providers and users alike.
Amy Copley is senior product manager of the Optical Switching division at Sycamore Networks (Chelmsford, MA).