White Paper

Next-generation SONET for Cable MSOs: Technical Overview

2

Next-generation SONET/SDHMSOs have several technology choicesavailable to prepare them for GbE-basedservice delivery and the inevitable growthin digital transport bandwidth. Thiswhite paper provides an overview ofNext-generation SONET digital transporttechnology choices available to CableMultiple System Operators (MSOs).

SONET standards were developed toprovide robust, point-to-point, TDM-based signal transport across short, medium, long, and ultra long haul optical networks. SONET standards are hierarchal in nature and were pre-dominantly designed for the transport of TDM-based voice signals. The tablebelow outlines the bit rates of hierarchalSONET-based TDM payloads.

The line side optical bit rate of SONETequipment has continued to expand tomeet growing voice transport require-ments. Over the years, the line side optical bit rate of SONET equipmenthas grown from 155 Mbps to 622 Mbpsto 2.5 Gbps to 10 Gbps. 40 GbpsSONET equipment is now emerging.Traditionally, OC-192 (10 Gbps)-basedSONET transport was predominantlyused in long haul applications in aneffort to pack as many TDM channelsonto a single optical carrier. Due to thereduction in the cost of OC-192 systems,this technology is now becoming cost-effective for metropolitan area applica-tions as a means to expand metro areadigital transport network bandwidth.

As SONET line rates have continued togrow, bandwidth and protocol flexibilityof most SONET equipment has failedto keep pace. This is becoming increas-ingly important in metropolitan areatransport networks where many differentnon-SONET data protocols and servicesnow need to be supported. In addition,it is now widely noted that data traffic

has overtaken TDM voice traffic ontoday’s digital transport networks.

To satisfy the need to support traditionalTDM applications as well as emergingnon-SONET data protocols such as GbEand Fibre Channel, extensive “next-gen-eration” standards development is andhas been taking place. The goals are toaccomplish the following:

• Provide SONET/SDH-compliantMAN and WAN transport of native,non-SONET protocols such as GbEand Fibre Channel

• Bring interface and protocol flexibilityto SONET/SDH transport equipment

• Greatly improve service granularity,density, flexibility, and bandwidthuse/efficiency for data-centric applications

• Drive intelligence into the SONET/SDH transport layer

• Continue to provide native delivery of traditional TDM payloads

• Deliver more profitable services byleveraging today’s installed SONETinfrastructure

Next-generation SONET/SDH solu-tions involve the implementation of sev-eral key, standards-based technologiesonto SONET/SDH-based transportequipment. These standards include:

• ITU-T G.7041 Generic FramingProcedure (GFP)

• ITU-T G.707/783 VirtualConcatenation (VCAT)

• ITU-T G.7042 Link CapacityAdjustment Scheme (LCAS)

• IEEE 802.17 Resilient Packet Ring (RPR)

Generic Framing Procedure (GFP)Generic Framing Procedure (GFP) is anemerging standard defined by ITU-TG.7041. GFP is a generic Layer 1 dataencapsulation mechanism designed toaccept and transport multiple native dataprotocols over metro and wide area digitaltransport networks. It has been designedto stop the proliferation of SONETmapping protocols (such as POS, x86

Table 1: SONET-based TDM payloads

Bit rate ANSI Asynch SONET

Name Tributary Transport

40 Gbps - - STS / OC-768

10 Gbps - - STS / OC-192

2.5 Gbps - - STS / OC-48

622 Mbps - - STS / OC-12

155 Mbps - - STS / OC-3

51 Mbps - - STS / OC-1

45 Mbps DS3 / T3 STS-1 SPE -

1.5 Mbps DS1 / T1 VT-1.5 -

64 kbps DS0 - -

3

LAPS, and proprietary methods) forpacket data services. GFP is deterministicwith low overhead providing a robust,high integrity, flexible mappingmethod for data protocols, such asGbE, into traditional digital transportnetwork payloads.

A common attribute of many widelyused, high-speed data protocols is theirreliance on 8B/10B physical layer cod-ing. 8B/10B coding involves the conver-sion of 8 bit bytes into 10 bit codewordsto ensure high transition density andDC balance (i.e. no long strings of 1’s or 0’s) for clock recovery circuits on thereceiving end.

Two types of GFP-based mapping havecurrently been defined. These include“transparent” (GFP-T) and “framed”(GFP-F). Each provides a unique set of advantages over the other, depend-ing on the application. The currentGFP standard provides a clean map-ping mechanism for emerging8B/10B data protocols, specifically GbE, Fibre Channel, ESCON, andFICON. Additional proposals arebeing considered to extend GFP-T for the support of DVB-ASI and FastEthernet (100bT).

Transparent GFP (GFP-T)Transparent GFP involves the map-ping of the complete client signal intofixed length GFP frames, regardless ofthe information content of the signal.It is intended for low latency trans-mission of client data signals wherethe inter-frame gaps contain impor-tant client specific data, such as sig-naling information or flow controlcharacters (e.g. Fibre Channel andESCON). If a significant amount

of the bandwidth in the client signalcontains needless idles (e.g. GbE inter-frame gap, preamble, start of framedelimiter), these idles are transported in a transparent fashionresulting in the consumption of unneces-sary transport network bandwidth.Finally, transparent GFP requires onlyminimal client signal awareness, therebyallowing a single hardware design totransparently transport multiple proto-cols based on 8B/10B coding.

Frame-based GFP (GFP-F)Frame-based GFP involves the mappingof client data frames only into GFPframes. Inter-frame bytes of the clientsignal are not mapped into GFP frames.Therefore, transport network bandwidthis better utilized, especially on lightlyloaded client signals, because meaning-

less idles and inter-frame gaps are nottransported across the network. WhileGbE, Fibre Channel, ESCON, andFICON share common 8B/10B physicallayer coding, they do not share commonframe delineation and packet formats. As a result, a protocol specific physicalcoding sublayer and MAC are requiredto extract and transport each protocol via frame-based GFP encapsulation.Therefore, different hardware designs arerequired to transport each of these protocols using frame-based GFP.

Figure 1 provides a pictorial view ofboth frame-based and transparent GFP encapsulation.

In conclusion, GFP-T and GFP-F offera differentiating set of capabilities andpermit a diverse range of data serviceofferings. GFP-T is ideal for Storage

SONET/SDH

STS-x-nv to client

Client input

Client PM

8B/10Bdecode

64B/65Bdecode

8B/10Bencode

64B/65Bdemap

GFP demap

GFP Core HeaderGFP payload area comprising 64/65 B blocks (octet aligned)— all client encapsulatedGFP-FCS

GFP map

Transparent GFPUsed for clients where the inter-frame gaps contains important client-specific information e.g. signaling information, flow control characters

SONET/SDH

STS-x-nv to client

Client input

Client PM

8B/10Bdecode

Remove all inter- frame bytes

8B/10Bencode

GFP demap

GFP Core HeaderGFP payload area comprising only client frames—not inter- frame bytes (octet aligned) GFP-FCS

GFP map

Framed GFPUsed for packet-oriented clients— no flow control or signaling charactersbetween packets

Replace all inter- frame bytes

Ethernet MAC frames, IP

Fibre channel, ESCON

Figure 1. Transparent GFP and Framed GFP

Area Network (SAN) applications whileGFP-F is better suited to Ethernet appli-cations. The following table comparesthese GFP framing formats.

Virtual Concatenation (VCAT)Virtual Concatenation (VCAT) is anexisting standard defined by ITU-TG.707/G.783 and ANSI T.105. VCATis not a new concept in principle. It wasoriginally designed to extend the utilityof the SONET/SDH transport layeralthough it hasn’t been adopted to dateas a mainstream networking technology.VCAT enables more efficient support ofpacket-based data services through the“right” sizing of SONET/SDH channelsand bandwidth.

SONET/SDH transport structures,including contiguous concatenation (i.e.OC-3c/12c/48c/192c), are optimized fortraditional TDM transport applicationsand are rigid in nature. VCAT provides a highly granular, flexible, and efficientmethod of provisioning traditionalSONET/SDH transport bandwidth fordata-oriented service transport. Speci-fically, VCAT allows for the grouping of non-consecutive SONET/SDH syn-chronous payload envelopes (SPEs) tocreate “virtual” concatenation bandwidthgroups. With VCAT, an arbitrary num-ber (X) of virtual containers (STS-1/3cSPEs) are grouped together with thecombined payload (STS-n-Xv) used tomatch the required bandwidth.

As an example, a wire speed GbE clientsignal (1,000 Mbps) can be mapped to aVCAT group of 21 STS-1s (21 x 51.8Mbps = 1088 Mbps) resulting in a band-width efficiency of approximately 92percent. Likewise, the same wire speed

GbE client signal (1,000 Mbps) can bemapped to a VCAT group of 7 STS-3cs(7 x 155.4Mbps = 1088 Mbps) toachieve the same bandwidth efficiency of92 percent. Conversely, if traditionalcontiguous concatenation is used for the same wire speed GbE client signal(1,000 Mbps), an OC-48c (2.488 Gbps)would be required, thereby wastingapproximately 60 percent of the provi-sioned bandwidth. The following tableshows how VCAT can provide substan-tially improved bandwidth efficiency ascompared to traditional SONET/SDH contiguous concatenation approaches for various data protocols.

As displayed, VCAT results in improvedSONET/SDH transport capacity effi-

ciency as well as much finer SONET/SDH bandwidth granularity. VCATcombined with GFP enables SONET/SDH transport elements to more effec-tively transport data-centric protocols,such as GbE, in their native formats.

Link Capacity AdjustmentScheme (LCAS)Link Capacity Adjustment Scheme(LCAS) is an emerging SONET/SDHstandard and is defined in ITU-TG.7042. As we have indicated, VCATprovides the ability to “right” sizeSONET/SDH channels and bandwidthresulting in more efficient support ofpacket-based data services. LCAS increasesthe flexibility of VCAT by allowing thedynamic reconfiguration of virtual concatenation groups.

4

Table 2: Characteristics of GFP-F and GFP-T

Table 3: Comparison of Contiguous Concatenation and Virtual Concatenation

Supported characteristic GFP-F GFP-T

Transparency to 8B10B—frame control codes no yes

At full rate minimizes WAN bandwidth (removes IFG) yes no

Permits per-frame performance monitoring yes no

Permits sub-rate bandwidth (with frame granularity) yes no

Minimizes latency for delay-sensitive protocols no yes

Long distance for Ethernet (local 802.3x PAUSE) yes no

Optional error correction for error-sensitive protocols no yes

Permits sharing of transport pipe among multiple clients yes yes*

Service protocol Client rate Contiguous concatenation Virtual concatenation

Fast Ethernet 100 Mbps STS-3c 65% STS-1-2v 98%

ESCON 160 Mbps STS-12c 26% STS-1-4v 78%

Fibre Channel 850 Mbps STS-48c 34% STS-3c-6v 92%

Gigabit Ethernet 1 Gbps STS-48c 40% STS-1-21v 92%

STS-3c-7v 92%

Fibre Channel 2 1.7 Gbps STS-48c 68% STS-3c-12v 92%

5

LCAS eliminates the slow and inefficientprovisioning process of classicalSONET/SDH networks and offers a means toincrementally enlarge or shrink the size of a SONET/SDH data pipe withoutaffecting the transported data. LCAS isused to add or remove members (STS-1sor STS-3cs) of a VCAT group resultingin the “hitless” provisioning of more/lessbandwidth over a live SONET/SDHVCAT circuit. LCAS uses a request/acknowledge mechanism that allows for the addition or removal of STS-1s/STS-3cs to/from a VCAT bandwidthgroup. Finally, the LCAS protocol worksunidirectionally, enabling service providersto provide asymmetric bandwidth. Thisis particularly important for MSOsrequiring asymmetric transport band-width as seen in VoD deploymentswhere content is distributed from VoDservers to edge QAM devices.

Figure 2 exhibits the advantages of GFP,VCAT, and LCAS. To employ GFP,VCAT, and LCAS over an existingSONET/SDH network, only the termi-nals at the end points of the connectionneed be modified. This allows serviceproviders to deploy GFP, VCAT, andLCAS in a simple manner by installingnew tributary cards. Likewise, they canscale these implementations by addingmore tributary cards to any other termi-nals in the network as required.

On a final note, GFP, VCAT, and LCAShave been added to the family ofSONET standards to improve SONETprotocol flexibility, bandwidth granu-larity, and efficiency. Nonetheless, theseimprovements do not change the circuit-based approach of SONET as the point-to-point nature of each data connectionremains. Additionally, protection band-width continues to be reserved for every

point-to-point connection and othernodes on the ring cannot claim unusedbandwidth. Resilient Packet Ring tech-nology addresses these concerns and isaimed at improving SONET/SDH net-work technology further.

Resilient Packet Ring (RPR)Ethernet is a low-cost, ubiquitous Layer2 transport protocol that dominatestoday’s LAN environments. The maindrawback of Ethernet is that it is apoint-to-point, unprotected data trans-port protocol that lacks carrier-classattributes. A router or switch processeseach data packet at every hop in anEthernet-based network. This attributecan be time consuming in larger Ethernetnetworks, thereby resulting in jitter andlatency performance that is typicallyinadequate for real time services such asvoice and video.

SDH/SONETDWDMOTIN

Client signal

Client signaling de-mapping

Payload multiplex

VCAT buffering(differential path delay)

STS-n-Xv

Payload inverse multiplex

Client-payload mapping

Interoperable global ITU-T standard— Mapping to SONET, SDH, and OTN

Figure 2. Features and advantages of GFP, VCAT, and LCAS

Generic Framing Procedure

• Efficient network resource utilization- Low GFP overhead- No over-provisioning with VCAT

• End-to-end service framing, PM and SP OAM, demarc

• Layer 2 independent- Supports RPR, other Layer 2 protocols

• Enables network consolidation- Converges next-generation services with existing

infrastructure investment

Virtual Concatenation

• Efficient network resource utilization- “Right-sized” SONET/SDH bandwidth channels- Reduction in over-provisioning- End-to-end proficiency for packet services

• TDM QoS, jitter performance- Deterministic QoS for all series

• Newsoft protection schemes- LCAS (ITU-TG.7042) allow dynamic reconfiguration

of VCAT payloads- LCAS provides automatic signaling capability

Legacy SONET/SDH networks offerclient access in the form of DS0, T1,DS3, OC-n, and in some cases, Ethernet.Today’s SONET/SDH networks offerexceptional jitter and latency perfor-mance and provide loss-less packet trans-port because packet data is transportedfrom source to destination without theneed for QoS decision-making inbetween. In addition, SONET/SDHnetworks reserve spare bandwidth capacity and provide 50 msec protectionswitching in the case of equipment faults or fiber cuts.

Resilient Packet Ring (RPR) is an emerging Layer 2 protocol that is beingdefined by the IEEE 802.17 workinggroup. RPR combines the low cost, efficiency, and simplicity of Ethernetwith the high availability, reach,resiliency, and scalability of SONETtransport technology. By combining the advantages of both SONET andEthernet, RPR provides support for new,connectionless, packet-based serviceswhile also providing traditional carrier-class features such as low latency and jit-ter, QoS, resiliency, and path restoration.

The result is the best of both worlds—a resilient, packet-oriented, ring-basedsolution that provides virtual mesh datanetworking connectivity.

RPR—Ring OperationResilient Packet Rings are ring-basednetworks that are optimized for connec-tionless data transport. As displayed inFigure 3, RPR networks employ at leasttwo counter-rotating physical rings used to connect all attached RPR network elements.

strip packets destined for them from thering thereby resulting in bandwidth con-sumption only on the traversed segment.Therefore, the unused spans on the ringremain idle and are available to otherstations and data streams. Spatial re-useis the term used to describe this propertyof RPR.

All attached nodes in an RPR ring sharethe available transport bandwidth with-out the need for provisioning circuits.The attached nodes automaticallynegotiate for access to ring bandwidthamong themselves via a fairness controlalgorithm. Additionally, RPR providesexpress treatment of all marked packets,thereby offering QoS for mission-criticaltraffic. Unlike Ethernet networks, RPRhas a transit path architecture that allows

Both physical rings are used to carryworking traffic, thereby allowing band-width traditionally set aside for SONETprotection to be utilized. This results in adoubling of available bandwidth capacity.For example, an OC-12 (622 Mbps)RPR ring provides 1.244 Gbps (2 x 622Mbps) of usable bandwidth.

RPR networks use a topology discoveryalgorithm such that all attached nodesautomatically learn the topology of thenetwork. In the topology map, each ele-ment stores a primary and secondarypath to every other element. Data is sentvia the optimal path only—typically theshortest path unless a network failurecondition exists. If a failure occurs, pack-ets are automatically sent to the destina-tion node via the secondary path within50 msec. Additionally, destination nodes

Outer ring data

Inner ring dataInner

ring control

Outer ring control

Figure 3. Dual counter rotating rings of RPR technology

6

7

packets to quickly bypass intermediatenodes en-route to their destination. Thisfeature allows RPR networks to providelow latency and jitter performance fortime-sensitive services such as voiceand video.

RPR Media Access Control (MAC)—details and featuresThe IEEE 802.17 working group is cur-rently defining the RPR MAC protocol.As with all Layer 2 protocols, RPRdefines a media access control (MAC)mechanism that defines the manner inwhich available bandwidth on the physi-cal media can be utilized by transmittingstations. MAC protocols, including theRPR MAC, also define how stationsshould react to congestion or collisionson the physical media. Finally, the RPRMAC prioritizes and buffers packets,thereby providing regulated access to themedia. Figure 4 provides a high-levelfunctional view of the RPR MAC.

As displayed, each RPR element containstwo MACs to provide communicationover the inner and outer RPR fiber rings.In a standard transaction, the host sys-tem sends control information, includingtopology, fairness, protection, and OAMstatistics, as well as data information tothe MAC control sublayer. The MACcontrol sublayer in turn sends RPRframes to the appropriate MAC fortransmission over the physical layer.

Packets received on either MAC are onlyrouted to the host client if a destinationaddress match is encountered. All otherpackets are placed back on the physicalring media to transit toward other nodeson the ring. This “transit” functionreduces packet delay significantly.

The RPR MAC offers four service classesto the host client. These include:

• Reserved—Unlike other services, theunused idle bandwidth for ‘reserved’services is not available to other ser-vices, thereby making this serviceclass similar to TDM circuits.

Interface to host system

RPR Layer 2 MAC defined by IEEE 802.17

Physical media such as SONET and Ethernet

Host c l ient

Physical Media Physical Media

RPR MAC

Media access controlMedia access control

MAC control sublayer

Control Data

Outer ring Inner ring

Figure 4. High-level RPR MAC functional view

• High Priority (Class A)—This service provides Committed Inform-ation Rate (CIR) and guaranteedbandwidth for high priority trafficthat is jitter and latency sensitive.Traffic in this class is not subject tothe RPR bandwidth-sharing algo-rithm. Voice, video, and circuit emu-lation applications are ideal for thisservice class.

• Medium Priority (Class B)—Thisservice is for CIR and other provi-sioned bandwidth applications requir-ing less stringent (but still bounded)jitter and latency requirements.Access to ring bandwidth is guaran-teed for this service class as well.Business data applications are idealfor this service class.

• Low Priority (Class C)—This ser-vice provides best effort access to ringbandwidth. RPR nodes use a band-width-sharing algorithm to negotiatefor a fair share of the ring capacity fortraffic in this class. RPR ring band-width is dynamically distributed for

8

traffic in this class. This service classis ideal for best effort, consumerInternet traffic.

The RPR MAC has several prime func-tional blocks that collectively allow it tooffer many advantages over legacyEthernet and SONET/SDH networks.These functional blocks include theMAC client interface, drop logic, transitpath, bandwidth manager, protection,topology, and physical layer as describedin Figure 5.

The purpose of each of the functionalblocks is as follows:

• MAC client interface—This inter-face allows the host to transmit andreceive data based on the service class.The RPR MAC client interface pro-vides the ability to back pressure thehost system, thereby ensuring thattransmitted packets have ring band-width available.

• Drop logic—The RPR MACinspects the destination address of allincoming packets to determine if thepacket should be stripped from thering or placed in the transit buffer for transmission to other nodes. The drop logic function performs the following functions:

- Strips uni-cast packets from thering and delivers them to theMAC client interface if a match-ing destination address to that ofthe station is observed

- Copies multi-cast packets to theMAC client interface and placesthem in the transit path for trans-mission to other nodes

- Places uni-cast packets in thetransit buffer if their destinationaddress does not match that ofthe station

- Strips bandwidth notificationpackets and passes them to thebandwidth manager

- Strips packets from the ring ifthey are corrupted or their time to live has run out

• Transit path—The transit pathallows packets to quickly bypass inter-mediate RPR nodes as they travel totheir destination. It contains a smallhigh-priority and large low-prioritybuffer that can hold up to two datapackets. Packets are stored herewhile waiting for the local host tocomplete its current transmission.The transit path is the key enablerof RPR’s ability to deliver reliable,time-sensitive services.

• Bandwidth manager—The RPRMAC tracks the bandwidth usage forlow priority packets and compares itto bandwidth notification messagesfrom downstream nodes. The systemdecides how much bandwidth isavailable for low priority data andwhether local host client back pres-sure should be applied. The band-width manager shapes and limits

bandwidth for each traffic class andenforces prenegotiated bandwidthlimits, as set by the RPR fairnessalgorithm, for low priority data. Thebandwidth manager also arbitrateswhether transiting packets from othernodes or traffic from the local hostwill be placed on the physical media.

• Protection—RPR provides a 50 msecprotection mechanism for packettraffic. It uses packet steering as adefault with the option for packetwrapping. With packet steering, allstations are notified of a networkfailure and transmitting stationschoose the ring that avoids the failedspan until they receive notificationthat the failure has been cleared.With packet wrapping, the nodesadjacent to the failure wrap traffic to the other fiber ring. Other nodeson the ring are notified and theyeventually steer their packets awayfrom the failure until notification that the failure has been cleared.

MAC client interface

Phys

ical

IF

H M L

Protection Topology

RPR MAC

Receive data

high medium low

Bandwidth manager

Drop logic Transit

Figure 5. Functional blocks of RPR MAC

9

• Topology—The topology discoveryalgorithm of RPR allows attachednodes to automatically learn thetopology of the network. Therefore,the host node can determine primaryand secondary paths to all otherattached nodes and thereby choosethe path a packet will take (typicallythe shortest path is chosen). Thetopology map is updated at regularintervals and after protection events.This process allows nodes to be addedor removed from an RPR ring with-out the requirement for a manualprovisioning process.

• Physical layer—RPR uses existingphysical layer solutions, includingexisting framers and optics. TheIEEE802.17 working group is devel-oping a reconciliation layer that willsupport PoS and GFP encapsulatedSONET and Ethernet physical layers.

RPR performance benefitsResilient Packet Ring technology offersmultiple performance benefits, including:

• Bandwidth efficiency—BecauseRPR uses both working and protec-tion bandwidth for live packet traffic,no bandwidth is wasted. Due to RPR’stransit path architecture and itsability to use bandwidth only alongthe transversed segment while strip-ping uni-cast packets at their destina-tion, “spatial reuse” of bandwidth isachieved. This results in furtherbandwidth multiplication becauseunused portions of ring bandwidthcan simultaneously be used betweenother nodes. RPR also provides ahighly efficient mechanism for trans-mitting multicast traffic. Unlikemeshed topologies where many packetcopies are sent over multiple paths,RPR nodes have the ability to receivea multicast packet as well as send it

on to the next node in the ring.Therefore, only one copy of a multi-cast packet is sent around the ring.

• Carrier-class packet service perfor-mance—With RPR, each node hastwo paths to every other node andpacket steering is used to automati-cally protect against fiber or equip-ment failures within 50msec. Inaddition, RPR provides exceptionallatency and jitter performance formission-critical applications such asVoice over IP.

• QoS and fair access to ring band-width—RPR provides multiple class-es of service for packets on or enter-ing the ring. New, differentiatedpacket services are easily accommo-dated over a common network byproviding separate classes forlatency/jitter sensitive traffic, com-mitted information rate, and besteffort traffic. The RPR fairness algo-rithm ensures that each node is givena guaranteed amount of bandwidth,thereby preventing any single nodefrom being starved from access toring bandwidth.

• Ease of management—RPR offersplug-and-play simplicity, negating theneed for manual provisioning whennodes are added or removed from thering. With automatic mechanisms fortopology discovery and protection,RPR requires little or no operatormanagement.

• Scalability—RPR networks providehigh scalability by allowing over 64nodes on a single ring. Nodes can beadded or removed from a ring withtopology discovery, bandwidth man-agement, and protection algorithmsautomatically accounting for thechange. Changes in bandwidth can beaccomplished quickly without com-plex provisioning steps or truck rolls.

RPR implementationsAlthough RPR can be used to replaceboth SONET and Ethernet networks, itis and will be used in combination withboth. For example, RPR is available onpresent day SONET/SDH transportequipment as well as on traditionalIP/Ethernet routing equipment.

RPR and the physical layerThe IEEE 802.17 working group is notdefining physical layers that are uniqueto the RPR Layer 2 MAC. Instead, theintent is to reference accepted andproven physical layers for use by RPR.Physical reconciliation sublayers are usedto translate between a physical layerinterface and the common, standardizedRPR Layer 2 MAC interface. The stan-dard currently defines such physical layerinterfaces for both SONET andEthernet physical layers.

For SONET/SDH physical interfaces,two reconciliation sublayers have beendefined. The SONET/SDH reconcilia-tion sublayer (SRS) defines an HDLC-like encapsulation for RPR frames that issimilar to PoS encapsulation. The GFPreconciliation sublayer (GRS) defines anencapsulation method using GenericFraming Procedure.

For Ethernet, the GbE reconciliationsublayer (GERS) defines an interfacebetween the common RPR Layer 2MAC, and the GbE media indepen-dent interface (GMII). The 10 GbEreconciliation sublayer (XGERS)defines interfaces between the RPRMAC and the 10 GbE media indepen-dent interface (XGMII) and 10 GbEAUI interface (XAUI).

10

RPR on SONET/SDH equipmentOn SONET/SDH optical network elements, RPR is implemented by integrating RPR-enabled Ethernet interface cards and packet switchingcapabilities to create an intelligent, distributed Layer 2 switch that usesSONET/SDH bandwidth as a virtualbackplane between end points. Figure 6 is an illustration of this approach.

Each RPR interface card is a highlyintelligent Layer 2 Ethernet switch.RPR cards come in multiple portdensities with various interfaces, such as 10/100 BASE-T, 10/100 BASE-FX,and GbE. They support standardEthernet protocols such as IEEE-802.1Q and IEEE 802.1p.

RPR-enabled SONET/SDH networksallow all, or a portion of, a ring’s totalSONET bandwidth to be provisioned as a ‘pool’ and dynamically distributedbetween multiple RPR interface cardsdeployed throughout the ring. Eachdynamic ‘pool’ of provisioned band-width is referred to as a virtual RPR ring as displayed in Figure 7. A singlenetwork can support multiple, indepen-dent, virtual RPR rings if required.

The key advantage of RPR-enabledSONET/SDH transport networks is theability to dynamically distribute band-width for packet traffic while continuingto leverage the non-RPR portion of network bandwidth for TDM services.Therefore, TDM services can be sup-ported in their traditional manner withno loss of performance as opposed toutilizing packet-based circuit emulationmethods. As a result, service providerscan use their currently installed SONET/SDH infrastructure to deliver newpacket-based, managed Ethernet servicescomplete with SLAs, while continuing to deliver legacy TDM services with noloss of performance.

Figure 6. RPR Implementation on SONET/SDH

RPR BW

TDM BW

Physical ring2 fibers—OC-12, OC-48 or OC-192

STM-3, STM-16, or STM-64c

RPR-based SONET/SDH NEs

RPR-based SONET/SDH NEs

Traditional SONET TDM ringDedicated bandwidth for circuit-based

connections such as DS1/DS3/OC-n

Virtual RPR ringDynamic bandwidth pool

for use by all RPR cards

RPR Ethernet cards provide distributed Layer 2 switching intelligence

Allocated RPR transport bandwidth is dynamically distributed between allRPR Ethernet cards, thereby providing

virtual mesh connectivity

Virtual IPR ringDynamic bandwidth pool

for use by all RPR cards

Figure 7. TDM and RPR transport on common SONET/SDH network

Resilient Packet Ring• Scalability• Reach• Robustness

SONET Distributed switch

Server

Server

Ethernet• Low cost• Simplicity• Universality

server

Server

Optical Node

Optical Node

Optical Node

Optical Node

11

RPR on IP routing/Ethernetswitching equipmentOn many IP routing/Ethernet switchingnetwork elements, Ethernet is used as aclient side interface and Packet overSONET (PoS) as a Layer 1 point-to-point WAN interface. Because RPR issupported on the SONET physical layerin a manner similar to PoS, it can beincorporated as a Layer 2 protocol intothe WAN side of today’s IP routed/Ethernet switched networks to providehighly resilient, virtual mesh connectivity.This is displayed in Figure 8.

Additionally, RPR will facilitate the dou-bling of traditional PoS bandwidth. Forexample, an IP-routed OC-48c POSring will have double the bandwidthcapacity (from 2.5 Gbps to 5 Gbps) with RPR while continuing to provide50 msec protection switching. Today,Layer 2 and Layer 3 network topologyinformation is exchanged between thesenetwork elements on a continuing basis.In the case of a network fault, Layer 2spanning tree protocol or Layer 3 pro-tocols, such as BGP and OSPF, must“re-converge” to route traffic around thenetwork fault. This process can, in manycases, take from several seconds to min-utes to complete. Because RPR provides50 msec protection switching, wheninstalled as a Layer 2 WAN interconnectmechanism between IP routing equip-ment and/or Ethernet switches, networkfaults can be avoided without triggeringtraditional Layer 2 and Layer 3 reconver-gence. With RPR, high layer protocolssuch BGP and OSPF are none the wiserthat a fault has even occurred.

In addition to increased resiliency, RPRemployed on the WAN side of today’s IP-routed/Ethernet-switched networkswill assist in significantly improving bothjitter and latency performance. This isdue to RPR’s shortest hop, transit-based architecture. With many oftoday’s IP/Ethernet or PoS physical ring-based architectures, a router or switchfully processes each packet through therouting or switching engine in order tomake a forwarding decision to the nextdevice. This results in higher packet loss,jitter, and unnecessary latency whencompared to RPR networks.

The outstanding resiliency, bandwidthefficiency, and jitter/delay performanceof RPR, combined with its QoS attrib-utes, make this technology especiallyimportant for MSOs wishing to use theirinstalled base of IP routing/Ethernetswitching equipment for the transport ofmission-critical traffic such as VoIP. RPRis poised to become a dominant Layer 2WAN interconnect mechanism for bothcore and aggregation IP-routed/Ethernet-switched devices within the MSO digitaltransport network.

IP router

IP router

IP router

IP router

RPR Layer 2 MAC implemented throughout ring on WAN side interfaces

Physical interfaceSONET/GbE/10GbE

Figure 8. RPR-enabled IP routed network

12

Next-generation SONETwrap-upAs we have discussed, GFP, VCAT, and LCAS are a collection of new andexisting standards designed to improveSONET/SDH protocol flexibility, band-width granularity, and efficiency. Thetable in Table 4 lists the standards thatdescribe these protocols.

It should be noted that these protocolsdo not change the point-to-point,circuit-based nature of traditionalSONET/SDH networks. These proto-cols are all Layer 1-specific. GFP, in both of its forms, is a Layer 1 frame to SONET mapping protocol. It is notdesigned to replace RPR, which defines a new Layer 2 MAC protocol. In fact,RPR can use GFP as a Layer 1 mappingprotocol and the IEEE 802.17 workinggroup is currently developing a reconcili-ation layer to provide this functionality.The GFP-based point-to-point datatransport approach and RPR ring-basedpacket data approach enable different,complimentary services as described byTable 5.

Finally, it is important to note that multiple pre-standard implementationsof RPR solutions have been on the mar-ket for some time. Specific examples ofthese include Nortel Networks OPTeraPacket Edge (OPE) and Cisco SystemsDynamic Packet Transport (DPT) solu-tions. However, the new RPR MAC as being defined by the IEEE 802.17Working Group fosters greater interoper-ability across both silicon and systemimplementations. The availability ofstandard-based products will benefitend users by supporting a multi-vendorenvironment. The IEEE 802.17Working Group was first formed inDecember 2001 and the first draft stan-dard was published in August 2002.

Table 4. Next-generation SONET standards table

Table 5. GFP (point-to-point) and RPR comparison

Point-to-point RPR Comments

Port aggregation x Point-to-point mapper: Just like any other TDM trib

RPR: Provides the ability to share bandwidth and aggregate to a single head-end port

Layer 2 capable x Point-to-point mapper: Provide transparent transport(logical customer of higher layer protocols. A front end Layer 2/L# device separation, CoS, etc.) would be required for customer separation.

RPR: Integrated Ethernet LAN

Subrate x x Point-to-point mapper: Provided by WLAN bandwidth management and flow control

RPR: Provided by token bucket rate limiting and flow control

Full rate over x Point-to-point mapper: Support for STS-24c, nxSTS-1 SONET (n=21) and nxSTS-3c (n=7)

RPR: Provided by WAN load sharing and 2.4G RPR

Spatial Re-use x Point-to-point mapper: No bandwidth sharing

RPR: Bandwidth shared between all stations on the WAN

Per port mappable x Point-to-point mapper: No bandwidth sharing

RPR: All Ethernet ports are shared

Fibre channel x Point-to-point mapper: USES GFP-T and a new trib

RPR: No support for PC

Region Standard Description Status

Global G.7041 Defines GFP. Approved2001 version + ‘02 addendum/corrigendum

G.707 (sect 11) Defines VCAT. Also defines Approvedcontiguous concatenation. 2000 version + ‘02

addendum/corrigendum

G.783 Detailed equipment specifications Approved(that include VCAT support requirements). 2000 version +

addendums

G.7042 Defines LCAS. Refers to G.707 and G.783 Approvedfor VCAT function. 2001 version + ‘02

addendum/corrigendum

G.709/G.798 Definition + equipment specification for VCAT of ODUk.

IEEE 802.17 Defines Resilient Packet Ring MAC. First draft for ballotQ3 2002.Expected final standardmid 2003.

ANSI T1.105 VCAT - Identical text to G.707/G.783 but Approvedwith SONET terminology. 2001 versionGFP, LCAS - References G.7041/G.7042.

ETSI EN300 417-9-1 Defines VCAT as per G.707/G.783. ApprovedNo recent updates to include GFP/LCAS.

13

Draft 2.0, code named “Darwin”, is themost recent version of the standard andthe first to be circulated to a full workinggroup ballot. The process will culminatewith the publishing of the standard,which in the case of IEEE 802.17 andRPR is expected later in 2003.

Nortel Networks offers MSOs a familyof market-leading next-generationSONET platforms that sets a new eco-nomic benchmark for cost reductionwhile providing new differentiated dataservices on existing networks. TheOPTera Metro 3000 MultiservicePlatform series features a fully non-blocking switching architecture provid-ing unmatched bandwidth managementcapabilities along with innovative servicemodules enabling the industry's highestdensity service termination withoutstacking multiple shelves. The OPTeraMetro 3000 Multiservice Platform seriesis equipped with Resilient Packet Ringtechnology that enables efficient band-width utilization with a full suite of nativerate client data interfaces. The OPTeraMetro 3000 series is deployed by majorMSOs and enterprises and is consistentlyrecognized as the market leader.

The OPTera Metro 3000 MultiservicePlatform series delivers increased net-work efficiency and service offerings by:

• Delivering significant improvementin Capex/Opex costs

• Providing optimal utilization of existing assets

• Enhancing service delivery • Enabling faster service deployment • Providing effective bandwidth

utilization

The Nortel Networks OPTera Metro3000 Multiservice Platform series offersunique packaging options and architec-tural advantages that significantly reducethe capital investment and lower theoperational expenses associated withbuilding and maintaining an optical network. The OPTera Metro 3000Multiservice Platform series providesforecast-tolerant and scalable optical networking solutions that enable serviceproviders to react to sudden changes in the marketplace without forklifting.These next-generation SONET solu-tions cost-effectively serve applicationsranging from customer premise tointer-office routes.

Nortel Networks is an industry leader and innovator focused on transforming how the world communicates and exchanges information. The company is supplying its service provider and enterprise customers with communications technology and infrastructure to enable value-added IP data, voice and multimedia services spanning Wireless Networks, Wireline Networks, EnterpriseNetworks, and Optical Networks. As a global company, Nortel Networks does business in more than150 countries. More information about Nortel Networks can be found on the Web at:

www.nortelnetworks.comFor more information, contact your Nortel Networks representative, or call 1-800-4 NORTEL or 1-800-466-7835 from anywhere in North America.

*Nortel Networks, the Nortel Networks logo, and the globemark design are trademarks of Nortel Networks. All other trademarks are the property of their owners.

Copyright © 2003 Nortel Networks. All rights reserved. Information in this document is subject to change without notice. Nortel Networks assumes no responsibility for any errors that may appear in this document.

GSA Schedule GS-35F-0140L1-888-GSA-NTEL

NN104002-050703

In the United States:Nortel Networks35 Davis Drive Research Triangle Park, NC 27709USA

In Canada:Nortel Networks8200 Dixie Road, Suite 100Brampton, Ontario L6T 5P6Canada

In Caribbean and Latin America:Nortel Networks 1500 Concorde TerraceSunrise, FL 33323USA

In Europe:Nortel NetworksMaidenhead Office ParkWestacott WayMaidenhead Berkshire SL6 3QHUK

In Asia:Nortel Networks 6/F Cityplaza 4,Taikooshing,12 Taikoo Wan Road,Hong Kong

List of acronymsAM Amplitude ModulationATM Asynchronous Transfer ModeBLSR Bi-directional Line Switched RingBW BandwidthCAP Control Access ProtocolCAPEX Capital ExpenditureCBR Constant Bit RateCMTS Cable Modem Termination SystemCO Central OfficeCRM Customer Relations ManagementDHUB Distribution HubDOCSIS Data Over Cable System

Interface SpecificationDPT Dynamic Packet TransportDVC Digital Video CompressionDVB-ASI Digital Video Broadcast

Asynchronous Serial InterfaceDWDM Dense Wavelength Division

MultiplexingETSI European Telecommunications

Standardization InstituteFR Frame RelayGbE Gigabit EthernetGbps Gigabits per secondGFP Generic Framing Procedure

HDT Host Digital TerminalHE Head-endHFC Hybrid Fiber CoaxialIEEE Institute of Electrical and

Electronic EngineersINM Integrated Network ManagementISP Internet Service ProviderIPPV Impulse Pay Per ViewIP Internet ProtocolJIT Just in TimeLAN Local Area NetworkMAN Metropolitan Area NetworkMbps Mega bits per secondMPLS Multi Protocol Label SwitchingMSO Multiple System OperatorNE Network ElementNEBS Network Equipment Building SystemNSG Network Services Gateway

Ethernet to QAM Converter madeby Harmonic Inc.

OAM+P Operations, Administration, Maintenance, and Provisioning

OC Optical CarrierPC Personal ComputerPCM Pulse Code ModulationPOS Packet over SONETPPV Pay Per View

PSTN Public Switched Telephone NetworkQAM Quadrature Amplitude ModulationQOS Quality of ServiceRF Radio FrequencyRPD Return Plant DemodulatorRPR Resilient Packet RingSAN Storage Area NetworkSDH Synchronous Digital HierarchySLA Service Level AgreementSNMP Simple Network Management

ProtocolSOHO Small Office, Home OfficeSONET Synchronous Optical NetworkSRP Spatial Re-use ProtocolSTB Set Top BoxSTM Synchronous Transfer ModeSTS Synchronous Transport SignalTD Transparent DomainTDI Transparent Domain IdentifierTDM Time Division MultiplexingUNI User to Network InterfaceVLAN Virtual Local Area NetworkVOD Video on DemandVoIP Voice Over Internet ProtocolVPN Virtual Private NetworkWAN Wide Area Network