Merging Circuits and Packets on One Network
Move over Sonet. Make way Ethernet. A new technology is raising the eyebrows of carriers and service providers alike that are looking for a packet-oriented, multi-service fiber optic network optimized for regional networks (LANs, MANs or WANs).
Combining the resiliency of Sonet with the price tag of Ethernet, Resilient Packet Ring (RPR) is an alternative Layer-2 technology that promises to cost-effectively and quickly provision new services on existing fiber. Hopes are also high that RPR's high degree of control and service assurance will help displace ATM technology.
Table 1: RPR networks combine the best of Sonet and Ethernet.
Courtesy of Resilient Packet Ring Alliance.
In basic terms, RPR is a set of switching nodes interconnected along a bi-directional, double fiber ring. Data is transmitted and load balanced on both rings. Unlike ring-based LAN technologies, RPR packets do not have to circulate the full ring, which effectively increases bandwidth availability and utilization. Also, RPR nodes need only "strip" the data packet to determine how the frame should be processed, speeding the delivery of services.
Expected to become a standard by Spring 2003, RPR is not intended for the Sonet-weary telco or carrier alone. As a matter of fact, some of the earliest interest and deployments are from MSOs looking for a single platform that leverages their current ring-based, fiber infrastructure to carry packet-based voice, video and data services.
"The driving force behind RPR is that existing Sonet and Ethernet technologies simply can't meet the emerging needs in the metro. The goal of RPR is to support voice and TDM-like services like Sonet, but still carry data very efficiently," says Robert Love, vice chair of the IEEE 802.17 Working Group and chair of the RPR Alliance.
"The beauty of RPR is that it can complement new technologies like 1 gigabit and 10 gigabit Ethernet rings and reduce the dependence on Sonet and its rigid provisioning and mapping of bandwidth through the network," says Tracey Vanik, technical director at RHK, a market research and consulting firm.
Even more importantly, RPR brings down equipment, facility and operational costs. Based on data from one vendor, Luminous Networks, RPR solutions de-layer the network and can consolidate network functions on one platform–leading to a 30 percent or more savings on equipment. And, with full, bi-directional ring utilization and dynamic bandwidth sharing, RPR maximizes current fiber deployments and can cut fiber costs in half.
Further equipment savings are realized from being able to interconnect one router to multiple routers using a single, cheap Ethernet port. For example, "an Ethernet port costs $200 versus current OC-3 ports that cost as much as $17,000," says Jay Shuler, VP of marketing and business development at Luminous.
Provisioning is another key area where costs fall. In doing away with Sonet's time-consuming circuit configurations, RPR allows one-layer provisioning. "Where RPR might take 20 actions to interconnect routers in a 20 node mesh, it would take an IP-Sonet mesh 570 actions. This is a 25-to-1 provisioning ratio, reducing operational costs by 10 times or more," adds Shuler.MSOs are thinking packets
As an MSO thinks about the future of its services, one word comes to mind: packets. Whether it's residential IP traffic, voice-over-IP or B-to-B services, packet-based services are the growth industry, and MSOs are looking for ways to accommodate them cheaply and easily.
"Going forward, we need to look at a transport box that switches on packets versus circuits and a single platform that supports voice, video and data," points out Greg Hardy, VP and GM of Scientific-Atlanta's Digital Transport Business Unit.
With its Prisma IP platform (a repackaged version of Luminous' PacketWave product), Scientific-Atlanta plans to serve MSOs and their CLEC arms with one solution that will eventually support interface cards for digital video, voice and CMTS.
Its partner, Luminous Networks, which has made a name for itself serving the telco market, considers itself the inventor of multi-service RPR. "Traditionally, Gigabit Ethernet is a data overlay over a circuit network, and what are called next-generation networks are really force fitting packets over circuit payloads," says Shuler of Luminous. "We have a converged infrastructure that runs circuits over packets, and our PacketWave product will eventually displace Sonet."
RPR may well complement Time Warner Cable's transition to Ethernet over IP. "We're looking at RPR as an extension to the Ethernet over IP protocol," says Dave Franklin, director of plant engineering. "RPR terminals are less expensive and easier to manage than ATM."
Currently trialing S-A and Luminous product, Franklin says that Time Warner looks forward to implementing a full standard product when it becomes available.RPR and the cable network
Oddly enough, the cable industry already relies on RPR. Some of the earliest deployments of RPR technology are in the backbone portion of the network by ISPs, such as ExciteAtHome.
Figure 1: Regional aggregation of data traffic from HFC networks onto a bi-directional packet ring.
Courtesy of Riverstone Networks Inc.
"It's very simple; the CMTS at the cable headend aggregates the data, which then connects to an RPR router on the packet ring and links up to the backbone network," explains Lewis Eatherton, architect of strategic engineering, ExciteAtHome. "Cable handles the HFC network; we handle the backend portion."
According to Riverstone, North American MSOs using packet rings via ExciteAtHome's backbone network include AT&T Broadband, Comcast, Cox, Rogers Cablesystems, Shaw Communications and Videon Cablesystems.
Riverstone Networks is courting the cable TV market with its RS 8000/8600 multi-service routers, and currently, has an RPR deployment with Telenet, Belgium's leading MSO. More than 52,000 kilometers of coaxial cable is linked by 8,500 kilometers of new fiber optic rings offering Internet access, IP-based virtual private networks and VoIP. The primary RPR ring has nodes in six major locations with 10 secondary RPR rings connecting 47 Telenet headends.
S-A's Hardy says two cable systems are now using Prisma IP. A trial system that relied on ATM is testing the technology, and another system is now running services over the new platform–allowing them to end a T-1 line leasing arrangement with a telco.
"While RPR and CMTS equipment usually reside at regional headends, Prisma IP (Luminous' PacketWave) and CMTS blades could also go into primary hubs, depending on the cable system," adds Hardy.
While no one is raising their hand to identify themselves, Luminous expects an announcement will be made with Scientific-Atlanta and a large MSO sometime in the fourth quarter of this year.
For MSOs, RPR might be the answer to reducing capital costs as new services get added to the mix: "RPR gets them circuits and packets on one network, reduces the cost of fiber and reduces operational costs since fewer platforms are needed and new services can be provisioned on the fly," says Schuler of Luminous.
Another cable-specific advantage to RPR is its multicasting capability. "Cable wants to broadcast content efficiently. On RPR rings, data packets are copied as they go from node to node, which is very efficient for multicasting," points out Tim Wu, technical marketing director at Riverstone Networks.Forging a new standard
Formed in November 2000, the IEEE 802.17 Working Group and its 126 voting members are charged with defining a new media access control (MAC) protocol for transporting IP data, packet video and voice over ring topologies with sub-50 millisecond protection switching–a carrier-class standard set by Sonet. However, RPR will be physical layer agnostic, which should speed up the standards-setting process and allow RPR to hook up effortlessly with Ethernet- and Sonet-based LANs.
To help with the promotion of the new standard and its marketing messages, an RPR Alliance was formed in January 2001. Principal members of the RPR Alliance include Alcatel, Alidian Networks, AuroraNetics, Ciena, Cisco Systems, Corrigent Systems, Dynarc, Lantern Communications, Luminous Networks, Mindspeed Technologies, NEC Corp., Nortel Networks, Riverstone Networks and Vitesse Semiconductor Corp.
"While the IEEE Working Group focuses on technical issues, the RPR Alliance can help when it comes to decisions that are more marketing-based and bridge differences," says Love.
While most standards efforts tend to be vendor-driven, Love points out that carrier involvement has been very high in this particular Working Group. Since the beginning of the standards process, carriers such as Worldcom, Sprint, SBC and ExciteAtHome have made presentations to the Working Group and helped define RPR requirements.
"Some of the key RPR requirements expressed by carriers revolve around sub 50 ms protection, Sonet/SDH performance monitoring and the support of both the Ethernet and Sonet physical layers," says Mannix O'Connor, director of business development at Lantern Communications, and secretary of the 802.17 Working Group and the RPR Alliance. Lantern Communications is a developer of advanced optical packet ring solutions for operators of MANs.
Figure 2: An RPR multi-service network can run circuits
and packets on one converged infrastructure.
Courtesy of Luminous Networks.
In September 2001, the Working Group debated final technical issues and started reading more thorough content and descriptions on the new standard–all an attempt to reach consensus as the standards move on to a first draft by November 2001. By spring of 2002, IEEE expects a single solution will be on the table for balloting, and a published standard should be complete by early 2003.
While 2003 seems a long way off, many find the timeframe typical for IEEE. In the meantime, pre-standard product is shipping. While there is disagreement on who has real RPR-like equipment, Cisco and Luminous seem to be the front-runners and are shipping product.
Many believe that pre-standard product is an important step to market acceptance. "It's always difficult to make products in the absence of standards, but you can't make standards in a vacuum," says Riverstone Networks' Wu. "With pre-standard product, customers can get familiar with the technology and run test beds."Make it a clean fight
Some contentious issues are still on the table. Besides technical issues, there are political and philosophical differences that need to be hashed out. Simply put, the debate pits Cisco against a group of much smaller companies on issues related to how the RPR standard should wind up.
Cisco, which developed the "de facto" standard known as Spatial Reuse Protocol (SRP), which is embodied in its Dynamic Packet Transport (DPT) product, would like to see the new RPR standard look very much like its current implementation. And who can blame Cisco? It has more than 170 customers and now has the luxury of directly dictating the features that get included in current chipsets.
But another camp has emerged in the IEEE Working Group. "The September meeting is a big turning point in the standards process. Dynarc, along with Luminous, Nortel, Alcatel, Lantern and Corrigent, is proposing a complete alternative to the Cisco solution," says Fredrik Hanell, VP of marketing for Dynarc.
Majority sentiment dictates that Cisco will not win this battle–and the company seems prepared for that eventuality: "We'd love for SRP to be adopted wholesale, but we see people are bent on fighting against it," says Jeff Baher, senior marketing manager for Cisco's Metro IP Access business unit.
And feelings do run high when it comes to Cisco's push for SRP. According to Vanik, "Cisco's DPT product can't even be considered a de facto standard. There has not been widespread adoption," says Vanik.
Perhaps summing it up even more succinctly: "We have deployed Cisco's DPT because we have no choice. It's either DPT or Sonet, and we're not interested in Sonet," says Eatherton of ExciteAtHome.
However, it's the very fact that RPR technology can break the Sonet "chains" that keeps everyone plugging away at a compromise.Making final technical decisions
Main areas of disagreement have to do with resiliency, an algorithm for controlling the order of packets on the ring and where the intelligence resides.
If a failure is detected along the RPR ring, some advocate a concept called ring wrap, while others, like Nortel, favor steering. "Since every RPR node should be aware of any failures and the status of all segments, the decision on which direction the packet should go on the ring should be made at the ingress point," says John Hawkins, marketing manager of Optical Ethernet at Nortel Networks.
Since RPR technology aims to do away with today's contention-based, best-effort methods for bandwidth allocation, an algorithm must be chosen that establishes the order of packets on the ring and their delivery.
One thing is clear. RPR will deliver packets on a fair basis, regardless of the customer's proximity to a node. "With RPR, carriers can now give due fairness for each customer which can then be translated to Service Level Agreements," says Vanik.
Another point of contention relates to the intelligence in the network. One side wants to rely on Layer 3 for intelligence, and others want it to be a customer decision.A living and breathing standard
In defining the RPR standard, the 802.17 Working Group wants a technology that can grow and meet future, unanticipated market needs: "There is no god standard," says Love, referring to the fact that there's no one standard that can anticipate every need. "We will define the MAC slice of the standard and pick up existing physical layers. But, we want people to be able to add-on to the standard so that new applications can be written," adds Love.
While the outcome of the standard is unknown at this time, the Working Group expects RPR standard product will be backwards-compatible with current versions in the field.
But for providers who decide to implement only a fully standard product, it might well be worth the wait: "This is true de-layering. With a single technology, voice and data can be managed on a packet network with tightly controlled latency and jitter," says Mannix of Lantern Communications.
Who could ask for anything more?