CONTENTS

Introduction Technology History Optical Ethernet Today Recent Trends Future Expectations

Introduction

LAN forms basis of all commercial,research,and data communication networks.Recently an increase in applications demand significantly higher band width. Ethernet is an easy to understand and extremely cost-effecti e technology. !or these reasons, "#$ of local area network %LAN& connections are now Ethernet based. 'o meet this,high speed LAN types ha e been de eloped ,including a number of ariations of basic ethernetLANs. ()*A+(,%carrier sense multiple access with collision detection& is used widely in wired LANs as a *A(%*edium Acess (ontrol& method. 'ransmission medium is coaxial cable %ether& up to -..km long,with repeaters at e ery .//meters.0pto -.1 machines could be attached to the system ia transcei ers screwed onto the cable.'he system ran at -."2*bps !our types of cablings are commonly used.'hey can be compared as shown.

Definition
4ptical Ethernet is the technology that extends Ethernet beyond the local-area network %LAN& and into metropolitan-area networks %*ANs& and wide-area networks %5ANs&. 5hile Ethernet LANs are almost exclusi ely used within the enterprise, optical Ethernet technology can be used as a ser ice pro ider offering. 6ey components of 4ptical Ethernet are the abilities to segregate traffic of different users and to deli er the particular ser ice le el each user purchases 'hey combine the flexibility, simplicity and cost effecti eness of Ethernet with the reliability, speed and reach of optics to allow users to extend their LAN en ironment across the *AN and 5AN.

Overvie
4ptical repeaters were part of the first Ethernet standard back in the early 3"#/s. 'oday, optical Ethernet ad ances promise to take Ethernet transport to le els undreamed of back then and not e en feasible using copper technologies today. 'hanks to ad ances in optical Ethernet, this most common %and most standard& of LAN technologies will soon be the most common %and most standard& of 5AN technologies. 'he history and potential of optical Ethernet technology is explained , focusing specifically on its impact on ser ice-pro ider networks and ser ices.
-

Technology
E thernet was defined as an open standard in the early 3"#/s by a consortium comprised of ,igital E7uipment (orp., 8ntel, and 9ER49: the resulting standard was called the ,89 Ethernet )pec. 'he goal was to promote a relati ely high- performance and low-cost LAN implementation using digital %baseband& signaling on a shared coaxial cable. 8n 3"#;, the 8EEE #/- Local-Area Network+ *etropolitan Area Network )tandards (ommittee %L*)(& released the #/-.; standard for Ethernet<a shared medium for LANs using a distributed media access control %*A(& mechanism called carrier sense multiple access with collision detection%()*A+(,&. Ethernet has pro en to be a highly flexible standard. 8t has e ol ed to include point-to-point signaling, full-duplex %unshared& links, and ery-highspeed networks. )e eral Ethernet characteristics ha e persisted, the most important of these being the framing format itself . An E thernet frame specifies a minimum and maximum packet si=e, a >protocol-type? field %expanded by #/-.; to function sometimes as a packet-length field&, and both a destination and a source *A( address. 'he address fields ha e been particularly aluable, as the globally uni7ue *A( addresses %each 2# bits long& ha e enabled the creation of new de ices %Ethernet switches& that ha e simplified the de elopment of large Ethernet networks.

4ptical links ha e been a part of Ethernet standards since the early 3"#/s. 'he needs and potential of optical transport, including point-to-point and full-duplex links, ha e dri en some of the e olution of Ethernet. 4ptical Ethernet technologies are pro iding the longest spans and greatest speeds used in LANs today, and they will undoubtedly, continue to do so in the future. 'he figure shown below re eals the frame format of ethernet .

Figure .Ethernet Frame

History The First Optical Ethernet Repeaters

'he first Ethernet standard included a pro ision for a single --km optical repeater in an Ethernet LAN that was expected to be used to link different buildings in a large LAN. As parts of a shared LAN, these links were not only half- duplex<they also propagated collisions %the signals used to limit access to a single sender at a time&. 'heir spans were limited by the maximum delay that could be allowed on an Ethernet LAN and still detect collisions. 'hey were true Layer-3 repeaters.

Campus Optical Ethernet

'he ad ent of the Ethernet >bridge,? now commonly called an Ethernet >switch,? changed the game. 'he purpose of an Ethernet bridge is to connect two different Ethernet LANs %the name >switch? e ol ed to denote interconnecting more than two&. 'his occurs at the *A( layer, Layer - of the @-layer 4)8 protocol model, and there are two important features in ol ed. !irst, not all traffic on either end is transported < only traffic destined for the >other? LAN. )econd, collisions %and collision-detection signals& are not transported: each side is its own Layer-3 LAN. 'ogether, these features not only impro e network performance by isolating LAN segments but also greatly increase the maximum si=e of an Ethernet LAN.
.

'he Ethernet bridge enabled large LANs to be deployed because a network of campus >bridges? could interconnect all of the building LANs. 8nstead of forming a simple >star? network, these were implemented as meshes, with multiple connections to each LAN. 'his re7uired the de elopment of the spanning-tree protocol %#/-.3,&, which works by disabling redundant paths and which implements a form of path protection for the LAN.

Aeople 7uickly reali=ed that if both ends of an optical link terminated on a bridge port, then the normal limits on the si=e of an Ethernet LAN segment no longer applied. 'he optical link could be operated as full-duplex, thereby doubling the bandwidth of the link. And with only one transmitter on a LAN segment, there could be no collisions: thus, the need to limit the si=e of the span for collision detection anished, allowing an optical Ethernet segment to span as far as the lasers could reach. 8n the early days, this still meant only a few kilometers because light emitting diodes and multimode fibers were used: but this was still enough to enable large campuses to be fully connected. 'he loss of collisions as a flow-control mechanism re7uired the de elopment of a new protocol, #/-.;x, to handle that need.

Optical Fast Ethernet

8n 3""., a new standard emerged for 3//Bmegabitsper-second %*bps& Ethernet transmission %or >!ast Ethernet?& o er (ategory-. %(at-.& unshielded twisted pair %0'A& copper cable. Actually, se eral standards were proposed and implemented, but only one gained significant acceptance<that was 3//CA)EB'9 using two (at-. 0'As at distances up to 3// meters. !ollowing close on the heels of the >copper? !ast Ethernet standards de elopment was the optical side. 'he first standards to emerge were adapted from fiber distributed data interface %!,,8& technology. 'he transcei er design and encoding techni7ue were the same, simplifying the work of the standards committee and ensuring that the standardi=ed technology would actually work. 5hile there were some differences, considerable effort was expended to make sure that !,,8 transcei ers could be readily adapted to optical !ast Ethernet. As on the copper side, se eral standards were ratified on the optical side. 'he first standard was for medium-range multimode fiber transmission at 3;3/ nm %3//CA)EB!9&, based upon the !,,8 standards. 'his pro ided for a normal range of about - km<ade7uate for most campus en ironments. 3//CA)EB!9 was part of the original #/-.;u !ast Ethernet specification back in 3""2. 'he second optical fast-Ethernet standard was 3//CA)E-)9, ratified in Dune of -/// %a full six years later&.
@

'his standard enabled backward compatibility with 3/CA)EB!L by using the same #./ nm wa elength and an auto-negotiation protocol. 'here was no single-mode fiber standard for optical !ast Ethernet transmission %and there is still none at the time of this writing, nor is one expected&. 'his lack of a formal standard has not stopped e7uipment manufacturers from implementing longhaul %3/ km B 3// km& fast-Ethernet links, and in practice they are likely to be interoperable, at least when operating at the same wa elength. 'hey are a ailable both at 3;3/ nm %the wa elength used in 3//CA)E-!9, !,,8, and )4NE' e7uipment& and 3../ nm %the wa elength used in wa elength di ision multiplexing E5,*F systems&. 'his de facto compatibility has resulted from the e olution of Ethernet de ices, which originally separated the digital logic from the analog logic.

'his distinction was formali=ed in the !ast Ethernet standard as the media independent interface %*88&. 'oday, the chip sets that handle Ethernet are generally independent of the chips that handle the media, be they copper, fiber, or potentially something else. *oreo er, the fiber-optic dri ers themsel es donGt know %or care& if the fiber is multimode or single-mode, what the type of connector is, or what wa elength is being used. 'he bottom line is that the approach used by Ethernet component manufacturers has led to a great deal of flexibility and interoperability, often transcending the standards that were created for that ery purpose.
#

Optical Gigabit Ethernet

'he #/-.;= gigabit Ethernet standard describes multiple optical specification. 3///CA)EB)9 describes short wa elength %#./ nm& transmission using multimode fiber with a maximum range of ../ meters on new fiber, or --/ meters on older fiber %with poorer dispersion characteristics&. 3///CA)EBL9 describes long-wa elength %3;3/ nm& transmission using either multimode fiber %with a range of ../ meters& or single-mode fiber %with a range of ./// meters&. As before, the standards committee took full ad antage of existing technology and >borrowed? the transcei ers and encoding formats of fiber channel. )pecifically, the !(B/ fiber dri er+recei er was copied, along with the !(B3 seriali=er+ deseriali=er. 'he #C+3/C encoding format was used, which specifies the framing and clock-reco ery mechanism. 'he significant changes were the signaling rate, which was increased to 3.-. gigabits per second %Hbps& from fiber channelGs 3./1 Hbps, and the frame content and si=e, which are the same as pre ious Ethernet implementations. *uch of the early gigabit Ethernet standards work was in specifying half-duplex operation in a collision detection en ironment e7ui alent to the shared LANs of ordinary and !ast Ethernet. Iowe er, that work was effecti ely wasted<all commercial Higabit Ethernet implementations today are point-to-point, full- duplex links.

"

Dust as in !ast Ethernet, the range of commercially a ailable optical Higabit Ethernet interfaces exceeds the limits outlined by 8EEE #/-.;=. 'here are de ices a ailable that operate at 3../ nm and de ices that operate at much greater distances than .km. 8n fact, 3./km spans are possible without repeaters or amplifiers. Again, this was enabled by the separation of Ethernet control logic from media control logic, which has now been formali=ed in a new way. 8t is not the formal definition of an *88 %called gigabit *88 EH*88F in the Higabit Ethernet standards&, but rather a new standard for a HigaCit 8nterface (on erter %HC8(&.

GBIC modules
4riginally specified for fiber channel, the HC8( standards ha e e ol ed to support Higabit Ethernet, and these modules ha e become the de facto standard for gigabit Ethernet interfaces. 'hey pro ide hot-swappable modules that can be installed to support LAN, campus area network %(AN&, *AN, and e en 5AN transport interchangeably, as needed. And they are a ailable from a large number of suppliers, keeping prices competiti e and capabilities expanding. 'he HC8( re olution is one of the best examples of industry consortia creating a new market by consensus, yielding a net increase in the re enues of all the participants. 'he modules themsel es, by irtue of
3/

being small and plugging into a standardi=ed slot, challenged the transcei er manufacturers, who responded

33

with technological inno ations and features beyond all reasonable expectations. 'he chip manufacturers contributed readily a ailable chip sets that support these HC8( modules, simplifying and speeding up the task of the hardware engineers.

Figure 2. Ethernet Timeline

Optical E t hernet Today

4p tical Ethernet systems are e ol ing beyond mere >optical links? that interconnect isolated LANs. Rather, they are becoming systems in themsel es, pro iding scale and functionality that is simply not feasible with copper-based Ethernet, including those linked by routers.

3-

LA
'o day, few optical Ethernet links are implemented within a computer room or small building. Cut there are exceptions for electrically noisy en ironments, highly secure transmissions , and ground isolation. And e en in a small building, it is easier to run fiber-optic conduit than electrical wires because there are fewer issues with building codes. 'raffic segregation is accomplished by using the 8EEE #/-.3pJ irtual LAN %KLAN& standard. 'his standard lets 4ptical Ethernet networks mark each userLs traffic with a KLAN tag as it enters the network and then use this tag to keep each userLs traffic separate as it crosses the network. 4f course, #/-.3pJ was designed for enterprise networks and the number of possible KLAN tags is too low. 5ork is under way in 8EEE to extend this number from 2,/"1 to approximately 31 million. 'his situation is likely to change, as ery-short-reach optics support much higher speeds than copper does. 8t is important to note that gigabit copper links are limited to about ;/ meters and that the next generation of Ethernet at 3/ Higabits would drop this already inade7uate distance dramatically.

CA
'oday, irtually all Ethernet links greater than -// meters are implemented optically. 'he (AN is dominated by multimode fiber, although most (ANs are really multiple LANs interconnected by routers that use optical links. 'his situation is changing, as the scale and functional

3;

capabilities of Ethernet switches increase. Ethernet 3/CA)EB' hubs, once dominant, no longer offer a sufficiently lower cost to Mustify their use. *ore and more LANs are being implemented with Ethernet switches, pro iding separate switch ports to e ery node on the LAN. 'raditionally, these Ethernet switches aggregate traffic into a high-speed uplink port %once Ethernet, then !ast Ethernet, and now Higabit Ethernet&, which feeds a router that itself interconnects the LANs and pro ides 5AN access. Cut today, when the cost of a router is weighed against the cost of a HC8( module, it loses e ery time. 'his trend is being accelerated by the proliferation of irtual LAN %KLAN&Bcapable Ethernet switches and by the de elopment of e en larger and more capable routers that understand and route to KLAN 8,s in the now larger LAN.

!A
4ptical Ethernet in the *AN is a relati ely recent de elopment. Higabit optical Ethernet has the capacity to pro ide direct Ethernet ser ices as a carrier offering, with ser ice switches that limit actual deli ered bandwidth as needed. *ultiple endors now offer direct Ethernet ser ices to subscribers, with only a few core routers linking those subscribers to the outside world %the 8nternet&. Cut >Ethernet ser ices? is not the leading reason to implement an Ethernet *AN today. Rather, the desire to reduce the number of routers in the network is becoming the most compelling reason to use a

32

Layer-- technology %Ethernet& in the metro area. 'he reason is simple< e ery router in a path, many of which are unnecessary, adds delay to the packet transport. 8n an ideal world, routers pro ide a buffer between management domains, which are generally companies or network pro iders. 'hey pro ide a place to control access, pro ide security % ia firewalls& and manage addresses. Cut within a management domain, layers of routers generate excess complexity and re7uire large staffs of >router guys.? 'hey also bypass a primary tenet of many routing protocols <that e ery router should ha e a direct link to e ery other router that it knows about. Cased fundamentally on Ethernet, 4ptical Ethernet *ANs let carriers deli er standard, well-known 3/+3//* bit+sec or 3H bit+sec Ethernet interfaces - the same as those used to easily connect office networks today. 8nstead of a )4NE' ring, the metropolitan backbone for 4ptical Ethernet networks will be based on the 3/H bit+sec Ethernet standard from the 8EEE working group #/-.;ae.

4ptical technologies enable Ethernet networks to extend o er much greater distances than campus Ethernet nets. Running o er single-mode fiber, 4ptical Ethernet lets links in the network range from ; to more than 1 miles in the case of 3;3/-nm wa elength technology, and up to 2;.2 miles for 3../-nm wa elength technology.

3.

'he routing network is much more effecti e %and easier to manage& if all of the entitiesG routers are directly interconnected, which is easily done today using optical Ethernet.

"A
Ethernet transport has not yet taken off in the long-haul network, but this is expected to change as 3/-Higabit Ethernet interfaces become a ailable. )ome of those are expected to operate at )4NE' 4(B 3"- speeds and at the distances needed for long-haul networks. 'he distance limitations are not a serious concern because most long-haul networks use dense wa elength di ision multiplexing %,5,*& systems to combine multiple
31

circuits o er a single fiber, each on its own wa elength, and these ,5,* systems pro ide the long-haul capability themsel es. speed, cost, and simplicity. (onsider a nationwide 3/-gigabit 8A )4NE'Bstyle ring implemented as 4(B3"- packet-o erB)4NE' %A4)& links between a do=en cities. Each city would need a large router with two relati ely expensi e A4) ports, and the a erage packet would tra erse half a do=en routers as it crossed the network. 8n the e ent of a link failure, the routers would spend a significant span of time con erging on a new set of routing tables to bypass the failure. Now consider the same network with each city containing an Ethernet switch with two 3/BHigabit Ethernet switch ports and a 3gigabit port connected to the local router. 'otal costs here would be much lower because 3/BHigabit Ethernet switch ports are expected to be much less expensi e than the e7ui alent router ports, and 3-gigabit router ports are relati ely inexpensi e. 'oday, se eral endors ha e products that aggregate and transport Ethernet traffic at 3/-gigabit speeds suitable for the 5AN. 'hese are proprietary, re7uiring matching de ices from the same endor at each end< normally not a problem in a full-duplex, point-to-point network.
Figure. LA #CA #!A #"A

)till, wide-area

Ethernet networks are expected to be implemented, the main reasons being

3@

Featured Products & Solutions: Ethernet over Fiber Ethernet o er !iber is primarily deployed in a point-topoint or mesh network topology, and deli ers packet ser ices o er dark fiber typically in the *etro Area Network %*AN&. Ethernet over RPR Ethernet o er RAR solution is through next-generation )4NE' products that enable simplification of network pro isioning, elimination of )4NE' bandwidth waste, impro ement of bandwidth efficiency, and link-layer protection. Ethernet over SONET/SDH Ethernet o er )4NE'+),I solutions deli er high-Jo) Ethernet Ari ate Line ser ices across an arbitrary multi-domain )4NE'+),I network, addressing established ser ice pro iders who want to de elop maximum return on network infrastructure assets.

3#

Ethernet over WDM Ethernet o er (oarse+,ense 5a e ,i ision *ultiplexing %5,*& is particularly effecti e in high-bandwidth, extreme-performance scenarios

Recent Trends Optical Ethernet $%itches

A true >all-optical network? is not likely to happen in the Ethernet space in the foreseeable future. 'he reason for this is that there are no technologies that enable packet switching at the optical layer today. Iowe er, purely optical Ethernet switches ha e already been demonstrated, in the sense that all of the Ethernet interface ports are optical %while the internal switching remains electrical&. 'his is todayGs technology optimi=ed for carriers selling Ethernet ser ices. 5hen a carrierGs customers are spread o er metropolitan distances %3/ km B ./ km&, then the lowest-cost ser ice-pro ider network is Ethernet with all of the de ice ports themsel es being optical. Cusinesses are re7uesting nati e Ethernet ser ices to interconnect their facilities into KLANs. 'his trend will be accelerated by

3"

the fiber-to-the-consumer mo e, as the lowest-cost optical ser ice will once again be Ethernet. )o the bulk of early fiber-to-the-home %!''I&, and possibly the maMority of fiber-to-the-business %!''C& ser ices, may well be optical Ethernet %at least as iewed by the carrier&.

GBIC !odules
'he latest HC8( inno ation is a miniBHC8(, also known as the small form-factor pluggable multisource agreement %)!A *)A& module. 'he miniBHC8( is only about half the si=e, effecti ely doubling the a ailable capacity that can be designed into the face of an e7uipment shelf. New designs are 7uickly taking ad antage of this space sa ings.

Resilient &ac'et Rings (R&R)

5hile not formally a part of Ethernet, the 8EEE #/-.3@ committee is creating a standard for packet transport o er fiber-optic rings. (urrent directions include using Ethernet framing in )4NE'Bstyle rings. 'he goal of RAR is simpleN to define a high-performance, high-a ailability optical transport suitable for carrier networks in metropolitan ser ice areas. Note that it is easier to implement a fast and robust link-failure reco ery in a ring topology than in a mesh topology: this is because in a ring, the alternate route is always known. 'he 8EEE #/-.3@ committee does not iew RAR as Ethernet, and indeed there is no intent that an 8EEE #/-.; de ice could

-/

be directly connected to an #/-.3@ interface. 'o a degree, a metro Ethernet is a competing mesh %not ring& technology with relati ely slow spanning-tree failo er, instead of )4NE'Bstyle, fast %./ ms& failure reco ery. )till, RAR implementations are likely to become a popular transport mechanism for Ethernet packets, especially for telecommunications ser ice pro iders

-3

*+,Gigabit Ethernet &roposed $tandards

'he 3/HEA %3/-Higabit Ethernet Alliance& is an industry consortium of about 3// members working to promote the acceptance and success of 3/-gigabit Ethernet. 'his group is not the same as the 8EEE #/-.;ae standards committee, which is working on a set of proposed standards for 3/-Higabit Ethernet. 'he manufacturers are not passi ely waiting on a standard, as multiple endors expect to announce 3/-Higabit Ethernet products during -//3, in ad ance of the standards ratification expected in *arch -//-. 4f course, each of these endors hope that their products will be one of the standards, and they promise that they will conform once those standard interfaces are defined.

*+,Gigabit Ethernet !a- Be Optical Onlhere are real challenges to making electrical signals carry a significant distance at those rates, and the implementations may simply be more expensi e than their more capable optical cousins %meaning that they simply wonGt happen&. 4ne of the proposed standards uses ery-short-reach optics, to be implemented as parallel data streams o er a fiber-optic ribbon containing 3- multimode fibers. 'his was proposed as a low-cost method to interconnect de ices in a room. A second proposed standard uses a ery compact package %about 3O x /.@.O x ;O& containing a coarse 5,* de ice, four recei ers, and four lasers operating approximately -. nm apart in wa elengths near 3;// nm. Each transmitter+recei er pair operates at ;.3-. gigabaud %data stream at -.. Hbps&. 'he proposed de ice has ery aggressi e engineering challenges and a ery aggressi e target-price point.

--

A third proposed standard is a serial interface using 12C+11C encoding %instead of the #C+3/C used in gigabit Ethernet&, a data stream of 3/./// Hbps, and a resulting clock rate of 3/.; Hbps. 'his is a fa orite of the Ethernet >purists,? who like the simplicity of Must mo ing the decimal place: moreo er, the resulting optical data stream may be directly interoperable with some of todayGs ,5,* systems.

A fourth proposed standard is a )4NE' 4(B3"- compatible stream, which is therefore clocked at ".".; Hbps. 'he disad antage is that it is not a pure 3/ times 3// megabit Ethernet. 'here may also be disad antages of cost<)4NE' is not the least expensi e transport method. 'he ad antages, howe er, are that it would be guaranteed to interoperate with all of the 4(B 3"- )4NE' de ices, including the networks of all of the maMor telephone companies, and all of the 4(B3"- compatible ,5,* systems. And the reality is that ".".; Hbps is close enough to 3/./// that no >real? applications are likely to notice the difference.

Figure . Band%idth,Gro%th Timeline

-;

Fu t ure E x pec t a t ions

Ethernet is %by a huge margin& the most successful networking technology e er. After all, "" percent of all '(A+8A packets %including 8nternet traffic& tra erse an Ethernet somewhere %and more likely fi e or six Ethernet LANs in the 5orld 5ide 5eb&. And the future is e en brighterP 'he Radiance 4ptical Ethernet )ystem is a new breed of optical access and connecti ity that pro ides a secure, yet manageable, Ethernet demarcation point between the ser ice pro ider and business user

Optical Ethernet to the Consumer

!''C is a reality for the !ortune .// today and will be a reality for ". percent of nonBsmall office+home office %)4I4& businesses within Must a few years. A carrier that runs fiber to a business enables the deli ery of all of todayGs communications ser ices and, likely, all of tomorrowGs. And Ethernet ser ices are the least expensi e ser ices that can be pro ided o er that fiber today.

!''I is already in trial deployments in se eral communities. )e eral studies ha e been completed suggesting that building an !''I network is no more expensi e than a full-scale ,)L buildout or a two-way cable tele ision %(A'K& upgrade, costing in the neighborhood of Q3,/// per home passed. 'he potential capacity of an !''I network is orders of magnitude greater than ,)L or (A'K, so it is clear that for any new buildout, laying fiber is the logical choice. And Must as in !''C, Ethernet is the least expensi e technology pro iding optical-network access today.

-2

8t seems clear that Ethernet ser ices will be coming soon to businesses near you< and to residences soon after. 'he only 7uestions are how soon, how fast, and how expensi eR 8n many metropolitan areas, some ser ice pro iders already pro ide Ethernet ser ices by the megabit of capacityN you can buy from 3 to 3// megabits, as needed A few ser ice pro iders already sell !ast Ethernet %3// megabit& 8nternet access for Q3,/// per month<a truly compelling price when compared to the 4ld 5orld )4NE'B or A'*Bbased ser ices. )till, is !ast Ethernet fast enoughR 4f course, that depends upon the business, but the cost of laying !''C dominates the capital expenditures, and todayGs !ast Ethernet infrastructure can be readily upgraded tomorrow. 'he switching cost and 8nternet-access cost are still much higher for Higabit speeds, but those costs are rapidly declining. !or Ethernet ser ices between locations for a business. Higabit Ethernet may already be needed to support file ser ers, backup ser ers, and other intranet applications. After all, most workstations installed today come with fast Ethernet built-in, implying that the switching and office-ser ices infrastructure should be significantly faster to a oid bottlenecking. Residential applications are more limited. 3/-*egabit Ethernet should be ade7uate for 5eb browsing for the near future, and the bandwidth needed for telephony is negligible. 'he most bandwidth-hungry application recogni=ed today is 'K, especially premium ser ices such as ideo-on-demand.
-.

A single channel or mo ie at broadcast 7uality re7uires about 2 *bps: ,K, 7uality re7uires only about " or 3/ *bps: and e en I,'K will probably only re7uire -/ *bps. 4f course, these figures are per-channeliewed, and the industry should plan on an a erage capacity of perhaps two channels per household at a time. )till, e en the most aggressi e of these numbers imply that a single !ast Ethernet ser ice to each home %not shared& is more than ade7uate for the ser ices we know of today.

Optical Ethernet Area

et%or's

Large optical Ethernet networks are changing the definition of the LAN. >Local? might e en be >global.? 'he original barriers in an Ethernet LAN %;-km span, 3/-; nodes, 3 optical repeater& ha e long since been an7uished. 'oday, the practical limits are more because of the need to terminate broadcast traffic or to pro ide security between management domains, or because of todayGs limits to the number of *A( addresses that an Ethernet switch might support. KLANs already start to address these issues: and larger KLANB enabled switches in the future are at least likely to control the problems and isolate them for a really large router to handle. 'he practical limits to the si=e of an optical Ethernet are not geographic: rather, they in ol e bandwidth, node count, and o erlying protocol %broadcast traffic, routing-table si=e, etc.&. As KLAN and other Ethernet ser ices become more common, we may e en see large corporate networks simplify into a single, optically connected Ethernet LAN, with only a few large routers pro iding the necessary functions of security, address management, and interdomain routing.
-1

'he largest optical Ethernet networks are likely to belong to carriers, interconnecting all of their points of presence %A4As&. E en todayGs optical Ethernet technology could be used to create a nationwide %or e en global&, ery- high-speed optical Ethernet that can be used to interconnect the large core routers in each A4A. )uch a network would ha e fewer router hops, faster link- failure reco ery, and lower cost than todayGs >normal? network of 4(B3"- A4) routers.

Be-ond *+ Gigabits
Dust as the growth of 3/-*egabit Ethernet led to the need for 3//-*egabit !ast Ethernet, and Must as the growth of !ast Ethernet led to the need for Higabit Ethernet, the growth of Higabit Ethernet is now dri ing the market to 3/-Higabit Ethernet. 'his trend is not likely to stop anytime soon. )er ers<whether 5eb ser ers, file ser ers, e-commerce ser ers, or others< must ha e greater bandwidth than the customers they ser e, otherwise, those customers will feel frustrated with inade7uate performance and possibly go elsewhere for ser ice. 'he best example is 5eb ser ers. 8 f the a erage 5eb browser is using a .1k modem, a ser er on a '3 line can simultaneously handle approximately ;/ customers. Cut the broadband mo ement has already started, and millions of consumers are now accessing the 8nternet from ,)L and cable networks. 'hese consumers access the 8nternet at speeds up to 3/-*egabit Ethernet %typical connecti ity for a cable-modem ser ice&, and for them, a ser ice pro ider limited to a '3 line is already unacceptably slow.

-@

'his re7uirement to always be faster than your customer %and by a factor of the number of simultaneous accesses& dri es the need for '; %2. megabits& or e en greater speeds today and will dri e the need for !ast Ethernet and Higabit Ethernet tomorrow. )ome carriers are already deploying gigabit opticalEthernet ser ices today. 'hey may limit their customers to a few megabits per second, but the links are gigabit- capable: and someday the fees for gigabit-scale Ethernet ser ices will be affordable. 'hen, e en a 3/-HigabitB Ethernet transport will be inade7uate. Engineers are already dreaming of the next step, and arguing o er what speed it will be. 2/-gigabit speeds %)4NE' 4(B@1#& ha e already been demonstrated, so that is a possible Ethernet target. Cut the Ethernet >purists? insist that the only logical next step is to mo e the decimal point one more time<to 3// gigabits. 8 n the meantime, the protocols and techni7ues for bandwidth sharing o er parallel links exist, work well, and are used in thousands of sites. 8t is a simple step to run parallel optical-Ethernet trunks, each on a separate wa elength, all multiplexed o er a single fiber pair using ,5,* technology. 8n this way, a point-to-point Ethernet link could ha e scores of 3/-gigabit channels, with an aggregate Ethernet bandwidth of perhaps 2// gigabits, todayP 0sing recently announced ,5,* capacity of 31/ wa elengths, 31//-gigabit-per-second links could be implemented. )o in a sense, 'erabit Ethernet is already a ailable<of course, this kind of network re7uires ery large Ethernet switches at the ends of that fiber.

-#

'he limits on optical-Ethernet bandwidth may Must be the limits of fiber-optic bandwidth< perhaps -. 'erabits per second for the a ailable spectrum on todayGs fiber, which is still well beyond the capabilities of todayGs lasers and electronics. )till, extrapolating from recent trends gets us to that le el in only . or 3/ years.

'he ad antages of optical EthernetN Ethernet is 3/ times less expensi e than the )4NE' technology being used today. Ethernet is a simple and widely understood technology. Ethernet is the best technology for carrying 8A traffic - Ethernet and 8A ha e grown up together. 4ptical Ethernet networks can easily handle the needs of both data and circuit-switched or oice applications. (ircuit traffic re7uires only modest bandwidth,

-"

REFERENCE

3.

www.iec.org www.optical-ethernet.com www.lightreading.com (omputer Networks - 'annenbaum A www.networkworld.com

-. ;. 2. ..

;/