A New Standard for MANs: The IEEE 802.17 Resilient Packet Ring
Over the years, a number of attempts have been made to develop a standard for a metropolitan area network (MAN) that would provide a high-speed packet-based data network capability similar to that available in the local area network (LAN), but extending over a larger area. The first serious effort, undertaken by the American National Standards Institute (ANSI), was the fiber-distributed data interface (FDDI), an optical ring technique operating at 100 Mbps. FDDI got off to a promising start but ultimately floundered, primarily because the equipment was expensive and FDDI didn't fit smoothly with existing LAN technologies. Later came the IEEE 802.6 distributed queue dual bus (DQDB) MAN standard. DQDB is based on the use of optical fiber bus technology, with data rates up to 155 Mbps. Again, for cost and compatibility reasons, DQDB never prospered.
Currently, two alternatives are available for supporting data traffic and for interconnecting LANs over a metropolitan area:
SONET (synchronous optical network) uses a dual optical ring topology that provides for rapid traffic restoration in the event of a link failure. However, SONET is inefficient because one of its two rings is reserved purely for failover.
An ATM (asynchronous transfer mode) network is made up of ATM switches and high-speed links. ATM provides high capacity and excellent QoS (quality of service) features. However, like SONET, ATM is considerably more expensive than comparable LAN equipment.
There has been much interest in extending Ethernet to the metropolitan area, using high-speed fiber links between Ethernet switches. The economic benefit is attractive: the ability to use inexpensive Ethernet hardware and phase out costlier WAN technologies such as SONET and ATM. Ethernet hardware is an order of magnitude less expensive than that required for SONET or ATM. However, Ethernet lacks the resilience of SONET and the QoS capability of ATM.
In response to the need for an Ethernet-like MAN solution, IEEE 802 has set up the IEEE 802.17 resilient packet ring (RPR) working group (http:/grouper.ieee.org/groups/802/rprsg) to develop a new standard for MANs using an optical fiber ring. IEEE 802.17 established five criteria that have guided the RPR work:
Broad market potential. There is already a large installed base of fiber-optic rings, but not optimized for data traffic. With the growing demand for IP-based data traffic in the metropolitan area, there is a potentially large demand for an effective MAN solution.
Compatibility. RPR must be compatible with existing IEEE 802 standards, especially those governing QoS and network management.
Distinct identity. No current effort within IEEE 802 addresses the requirements of a high-data rate, resilient MAN to provide QoS for a variety of traffic types.
Technical feasibility. Implementations of candidate proposals for an RPR already exist, demonstrating the feasibility of this general approach.
Economic feasibility. Fiber-optic and related costs are such that the RPR will provide a cost-effective solution to MAN requirements.
IEEE 802.17 Physical Layer
As with any IEEE 802 standard, IEEE 802.17 will include both a physical and a medium access control (MAC) layer. The essential feature of the physical layer is the use of a dual-ring topology using optical fiber at high data rates, up to 1 Gbps or more. Under normal operation, data can be transmitted simultaneously on both rings, doubling the capacity. The dual-ring topology provides robustness by including a capability for automatic reconfiguration after a link failure. There is already a massive deployment of fiber-optic rings in MANs. These rings are currently using protocols that don't scale to the demands of packet networks. Thus, IEEE 802.17 leverages existing infrastructure to meet data transport needs.
The ring topology provides a number of benefits:
Efficient resiliency. The dual-ring topology provides robustness by including the capability of automatic reconfiguration after a link failure (see Figure 1).
Decreased number of ports. Distributed-ring networks need half the number of ports to provide the same amount of resiliency as centralized switched networks with redundant paths. The result is decreased investment and maintenance costs.
Scalability and step-by-step network rollout. More routers can be added to the ring incrementally, adding packet-forwarding capacity to the ring as the customer base grows. A key to the ease of scalability is the nature of the RPR data handling. Each node on the ring offloads frames addressed to the local station and simply passes through other traffic. Thus, the full traffic need not pass through the router, but merely through an interface card that attaches the router to the ring.
High bandwidth utilization. Through the use of spatial reuse algorithms, the bandwidth can be reused several times on the ring.
Simplified maintenance. In distributed packet forwarding, there is no need for costly centralized routers, limiting the risk of forklift upgrades (a wholesale changeover of hardware).
Simplified traffic forecasting. Optical link resources are automatically distributed among the nodes of the ring and thus traffic forecasting is calculated on an aggregated level, without the need to predefine capacity between the routers that are connected to the ring.
Inherent multicast support. One of the major advantages of using shared media structures is the simple and straightforward support for multicast.
Figure 1. Resilient packet ring (RPR) operation.
The actual transmission scheme used at the physical layer is flexible. RPR is designed to operate over SONET, wavelength-division multiplexed (WDM) fiber, or Ethernet-defined fiber links.