A fresh approach for the metro: RPR present a complementary approach to myriad networking technologies – Network Edge – resilient packet ring

A fresh approach for the metro: RPR present a complementary approach to myriad networking technologies – Network Edge – resilient packet ring – Brief Article

Jay Shuler

Carriers today are looking at a range of technologies to relieve the metro bottleneck, including RPR (resilient packet ring), which is currently being standardized by the IEEE 802.17 working group. So what exactly are the relationships among RPR and various other networking technologies-SONET/SDH, MPLS, Ethernet switching, DWDM, voice and TDM? Are they complementary or redundant?

RPR is a MAC (Media Access Control) protocol closely related to Ethernet but designed to optimize bandwidth utilization and facilitate services over a ring network. It is designed to provide the carrier-class attributes normally associated with SONET and SDH, including 50-ms ring protection, full FCAPS management and TDM service support over an infrastructure optimized for emerging packet services (see Figure 1). RPR then becomes a multiservice transport protocol based on packets rather than circuits, and RPR systems are viewed as the logical successor to SONET/SDH ADMs (add/drop multiplexers). RPR provides both legacy TDM and differentiated packet-based services (such as native Ethernet) over a single, converged network. For carriers, it promises to deliver all the needed end-user services-TDM voice, VPN data and Internet access-at dramatically lower equipment, facility and operating costs. TDM voice is typically delivered over synchronous T1/El or DS3/E3 circuits, which can be carried over RPR without compr omising toll quality. VPN and Internet data services-today delivered using PPP (Point to Point Protocol), frame relay, or ATM over TDM circuits (all the way from DS1 to OC-48)-can also be carried over an RPR-based transport infrastructure in their native form. Furthermore, these services can gradually migrate over an RPR-based infrastructure and be carried more efficiently using native Ethernet private lines or TLS (transparent LAN services) with QoS parameters equivalent to frame relay.

Locations with plenty of fiber can build mesh networks and may use other approaches for protection. These networks use point-to-point links for interconnecting nodes. This topology assumes an abundance of fiber and an ability to restore service from any fault in sub-50-ms timeframes. Algorithms for this purpose are rare and often proprietary. With the RPR MAC in ring topologies, fast protection and high ring utilization are relatively straightforward propositions.

Carrier Services and Costs

Service providers are interested in technologies that allow them to deliver data and voice services profitably. Infrastructure decisions require the careful balancing of revenue from services offered, initial infrastructure costs, and ongoing operational costs. While data services are growing rapidly, they are frequently low profit offerings. RPR is being designed to address this fundamental issue. The highest margin services are those that have a private line-like behavior. Traditionally, that type of service was only available on a SONET infrastructure. With the advent of RPR, carriers can now offer SLA-guaranteed services on an infrastructure optimized for data or packet services. This will enable higher margins on data services and is an essential prerequisite to additional spending for improved infrastructure.

RPR over SONET/SDH and Ethernet

RPR is physical-layer-(L1)-independent, meaning that it can operate over the Gigabit Ethernet physical layer (8B/10B), the new 10-Gbps Ethernet (10 GigE) physical layer (64B/66B), or a standard SONET/SDH framing layer. This gives vendors and carriers the flexibility to choose RPR implementations to meet different networking applications. For example, a new network build can make use of 1-Gbps, 2.5-Gbps, or 10-Gbps RPR over an Ethernet physical layer, while the same carrier can implement RPR over SONET in another location.

RPR, however, will not channelize bandwidth inside SONET sub-channels such as VT (virtual tributary), STS, or STM. RPR dynamically allocates service bandwidth within the entire available bandwidth of the facility, whether SONET or GigE.

Interoperating RPR and MPLS

RPR and MPLS serve very different and complementary purposes. MPLS is a control plane that provides a virtual connection, edge-to-edge, across a heterogeneous network of switches and routers. It is used to maintain customer and service class separation of traffic, provides an entity that can be measured and billed, and allows carriers to engineer their traffic to maximize the efficiency and performance of their networks. An MPLS LSP (label-switched path) can be thought of as defining a chain of links across the network, while RPR defines the links in that chain. RPR’s role is to provide rapid healing of the links in the MPLS chain so that the MPLS service remains undisturbed.

Some vendors propose using MPLS to provide rapid recovery of connections. To understand the issues associated with this, it is necessary to understand that protection switching involves four steps:

* detection;

* notification;

* calculation; and

* protection switching.

To provide protection-switching speeds that are adequate to maintain the integrity of TDM circuits such as T1, all four of these actions must occur within 50 ms. Because of the length of an MPLS chain, detection and notification cannot be guaranteed to occur within this timeframe. Furthermore, because MPLS typically functions across many rings and spans in a mesh network, the calculation of a protection path is a complex and time-consuming process that cannot be performed in 50 ms. To overcome the issues associated with protection of an end-to-end path, some vendors have even proposed some local path recovery schemes, but these tend to be complex and cumbersome to manage, because alternate routes for each individual path have to be preconfigured at various points of failures in the network. With RPR, the link is repaired before the MPLS chain even knows it’s broken.

RPR Used with DWDM

It is sometimes argued that DWDM is the current approach for metro service delivery. RPR proponents agree that DWDM will play an important role and the technologies complement each other in most metro designs. RPR will act as an aggregation network, taking dozens or hundreds of high-speed connections into a fiber ring. It will multiplex these signals to insure QoS and efficient bandwidth utilization. At some point, however, the 2.5-Gbps or 10-Gbps RPR ring will fill and more bandwidth will be required. At this point, the introduction of DWDM into the metro access network makes sense if it is less expensive than running new fiber rings. However, in many cases it is not possible to run new fiber. This is a common instance where the bandwidth on existing fiber rings can be increased using WDM to gain more capacity.

CWDM (Coarse WDM) and other fiber multiplexing approaches may also feed RPR rings from customer premises locations. For locations where fiber constraint exists between the customer and the RPR ring, CWDM is a solution for running multiple customers and very high bandwidth over the existing point-to-point link.

RPR and Ethernet Switching

Most proponents of RPR expect carriers to offer Ethernet services over the RPR infrastructure. Since most enterprise locations use Ethernet as their internal networking protocol, carriers should gain an advantage by delivering an Ethernet port to the business location. This enables customers to connect seamlessly to the carrier infrastructure. As carriers deploy RPR, they will be able to offer higher bandwidth connections to their subscribers. Typical access rates will move from T1 rates of 1.54 Mbps to 10 Mbps, 100 Mbps, or even Gigabit data rates. SLAs are crucial to the ubiquitous deployment of high-speed Ethernet services. These QoS guarantees and SLA contracts are assured with RPR because the RPR MAC will implement an access control algorithm that is designed specifically for metro carrier networks. This capability will drive even greater deployment of Ethernet switching in the enterprise.

Ethernet as a networking technology is deployed increasingly in other devices and in wireless applications. The low latency, jitter and QoS guarantees of an RPR backbone will improve QoS at the network edge. Lower cost and better QoS will enable more Ethernet applications at the edge of the MAN/WAN network.

Voice and TDM Traffic

RPR is designed to enable service providers to build carrier-class, data-optimized networks. However much revenue today comes from TDM services over legacy TDM networks. Circuit emulation of TDM services over RPR-based networks enables carriers to maintain their current high-revenue TDM services as well as offer new Ethernet services.

Because TDM services are delay and delay-variation sensitive, circuit emulation demands strict jitter and wander control as well as end-to-end synchronization over the ring. Standardization of TDM circuit emulation over generic packet networks has been progressing in the IETF. Currently, standardization is focused on circuit emulation services over MPLS. MPLS labels and a new circuit emulation header are used to encapsulate TDM signals and provide the circuit emulation service over MPLS (CEM).

Until circuit emulation of TDM over MPLS is fully standardized and interoperable, aggregation of TDM traffic headed outside of the metro ring can be done at a central point and handed over to a long-haul TDM network.

Current Deployments

Although RPR is not yet standardized, at least three prestandard variations of it are already being put into live service in Asia, Canada, Europe and the United States. Many carriers, including established ILECs and IXCs, are watching RPR developments carefully and planning for wide-scale adoption as soon as standards-based products are ready. Most of the major carriers have actively participated in the IEEE 802.17 Working Group, articulating their requirements and taking a keen interest in the evolution of the standard.

Jay Shuler, Luminous Networks; Mannix O’Connor, Lantern Communications; and Nigel Cole, Corrigent Systems, are members of the RPR Alliance. (www.rpralliance.org)

COPYRIGHT 2002 Horizon House Publications, Inc.

COPYRIGHT 2002 Gale Group