With the proliferation of Web Services, it’s not just the number of packets traversing the network, but the size of those packets. XML—one of the emerging protocols that enable Web Services—is particularly “verbose”. Kobielus2 estimates that XML requires 2 to 100 times more bits than non-XML binary transfers. XML messages are bulky and will put a strenuous load on networks.
When applications execute on a single server, inter-process communications are handled by the server’s system bus. When processing is distributed between locations with Web Services, interprocess communications are carried over the network. Since network performance is never at the level of server system bus performance, special care must be taken when integrating Web Services into a business process. Having a network that behaves like a system bus— that is, predictable, deterministic and low latency—simplifies that integration task. And it allows programmers to have tighter time-out windows for Web Services requests, which can dramatically improve total applications performance, and ultimately the end-user experience. So low latency is an increasingly critical network characteristic.
As Web Services proliferate within an organization, applications will become increasingly interconnected in often unpredictable ways. They will become very difficult to distinguish when a specific communication is mission-critical. For example, assume a brokerage house has an arbitrage program that trades on real-time market fluctuations. All packets associated with that arbitrage program would be considered high priority. Now, let’s suppose that arbitrage program trades crop futures.
White PaperAdaptive WANEnabling Web Services, Networked Remote Storage and Grid ComputingThree major developments are significantly affecting Wide AreaNetworks (WANs): Service-Oriented Architecture (SOA), or Web Services a fundamentalshift in the way enterprise software is being implemented. Networked Remote Storage organizations increasingly need access to geographically separated storage assets, whether to integrate amerger/acquisition, to satisfy Business Continuity/Disaster Recovery(BC/DR), for competitive advantage, or simply to comply with newregulatory and legal requirements. Grid Computing the latest in information technology (IT) asset virtualization.These three tremors on the IT landscape (Web Services, NetworkedRemote Storage and Grid Computing) individually and collectively,are triggering exponential growth in inter-site traffic.Organizations will need to react to this exponential growth. If theyreact as they have in the past adding bandwidth incrementally andprioritizing high priority traffic during peak traffic periods organizations will find their applications increasingly struggling andcompeting for congested WAN resources. Their investments in IThardware and software will not realize the expected benefits. Whilethis approach may have worked before, when bandwidth needs weregrowing linearly, a more proactive plan is necessary to addressexponential traffic growth.Being more proactive means moving from multiplededicated single-application networks to a moreflexible, adaptable network architecture what Cienarefers to as Adaptive WAN. Adaptive WANs providehigh throughput, low latency connectivity to meetnetwork requirements of emerging IT applicationsover any geography. In metropolitan areas wherefiber is available and economical organizations willincreasingly move from leased connectivity to privatefiber-based networks. Over longer distances, where fiber is notusually an economic choice, enterprises will increasingly need to relyon networking techniques where high bandwidth and low latencyhave been key considerations. Until recently, these types oftechniques have been used mainly for storage system connectivity.Web ServicesWEB SERVICES ARE DISTRIBUTED, OBJECT ORIENTED PROGRAMSthat can work together across distributed computing platforms.Applications developed with Web Services will have a very significantimpact on WANs. Why then are enterprises embracing WebServices? Web Services enable organizations to easily access information fromheterogeneous systems. Rather than writing expensive gateway programsbetween applications, interoperation is achieved through Web Services.Organizations also can avoid debugging the gateway whenever softwareis updated on either side. Web Services also promise more rapid applications development. Bypatching together Web Services into a collective application, enterprisescan get to market faster with new business processes a clearcompetitive advantage. Software development is more economical with Web Services. Developers can focus on their value-add and re-use software that hasbeen written before.Traditional ApplicationWeb ServicesApp #1App #2App #NApplication #1 Application #2 Application #NSharedComponentsXML interfaceFigure 1. Software Building Block ReuseExecutive SummaryUntitled Document2A recent survey by Kubernan, indicates that almost 90% oforganizations will employ Web Services within the next year, and twothirds of organizations will significantly use Web Services. This iswithout a doubt, the key software technology of this decade.Web Services can be flexibly deployed over any CPU, in any device.This new distributed computing model enables applications to bedeployed as services accessed by anyone inside or outside theenterprise with only a web browser and suitable plug-ins. To actuallymake such new applications work requires a new underlying applicationinfrastructure, known as an SOA.Initially envisioned to run across Local Area Networks (LANs) within adata center SOAs also play a critical role in enabling an evengreater return for the enterprise by allowing Web Services and othermission critical applications to be run between multiple sites. SOAsrequire the same attributes for performance response-time andavailability for the WAN that are found in the LAN.Applications have, until now, executed mainly on a dedicated serveror cluster of servers. When applications are patched together fromWeb Services, parts are executed here and parts are executed thereand other parts are executed elsewhere. All those disparate partsmust communicate with one another. With most organizations, it snot very likely that those disparate parts are at the same physicallocation. So the result is a lot of inter-site traffic. One of the drivers of Web Services is to operate business processes more quickly. Thiscreates improved processes which will, in turn, require even moreinter-site traffic. Some analysts predict that, XML traffic will triple onenterprise networks by 2008, accounting for close to 50% of themessages sent. 1With the proliferation of Web Services, it s not just the number ofpackets traversing the network, but the size of those packets. XML one of the emerging protocols that enable Web Services isparticularly verbose . Kobielus2estimates that XML requires 2 to100 times more bits than non-XML binary transfers. XML messagesare bulky and will put a strenuous load on networks.When applications execute on a single server, inter-processcommunications are handled by the server s system bus. Whenprocessing is distributed between locations with Web Services, inter-process communications are carried over the network. Since networkperformance is never at the level of server system bus performance,special care must be taken when integrating Web Services into abusiness process. Having a network that behaves like a system bus that is, predictable, deterministic and low latency simplifies thatintegration task. And it allows programmers to have tighter time-outwindows for Web Services requests, which can dramatically improvetotal applications performance, and ultimately the end-userexperience. So low latency is an increasingly critical networkcharacteristic.As Web Services proliferate within an organization, applications willbecome increasingly interconnected in often unpredictable ways.They will become very difficult to distinguish when a specificcommunication is mission-critical. For example, assume a brokeragehouse has an arbitrage program that trades on real-time marketfluctuations. All packets associated with that arbitrage programwould be considered high priority. Now, let s suppose that arbitrageprogram trades crop futures. It s not unreasonable to have a scenarioAdaptive WAN: Enabling Networked Remote Storage and Grid Computing1ZapThink senior analyst Ronald Schmelzer, November 20042 Wrestling XML Down to Size , James Kobielus, BCR, December 2004 We currently make extensive use of Web Services We currently make only moderate use of Web Services, but it is likely that we will significantly expand our usage We currently make only moderate use of Web Services, and that it is unlikely to change We don t currently use Web Services, but it is highly likely that we will in the next 12 months We don t currently use Web Services, nor do we intend to in the next 12 monthsWeb Services UsagePercentage ofRespondents33.7%32.0%18.6%5.2%10.5%Table 1. Employment ServicesSource: KubemanUntitled Document3where a crop futures application calls a weather service. Normally aweather service s packets would not be prioritized, yet there s ascenario where weather results can be in the critical path of croparbitrage programs. A situation could easily arise where so-calledlow priority weather packets are bounced to make room for arbitragepackets at the very moment the high priority crop arbitrage programis awaiting weather results.Networked Remote StorageIncreasingly, organizations are struggling with Networked RemoteStorage. Organizations have always been spread out whether forproximity to customers or proximity to suppliers. Recent mergers andacquisitions have given organizations even more data center spread across geographies. Additionally, BC/DR concerns havecaused organizations to locate data centers further apart. At thesame time, there are organizations engaging in follow-the-sun operations where information is transferred from continent tocontinent as the working day evolves from Asia to Europe then to theAmericas and then starting all over again. So, organizations arebecoming reliant on storage systems that must be interconnectedacross wide geographic distances.Not only are storage assets distributed geographically, but the abilityto access the information they contain is becoming more important.Enterprises must now respond to a myriad of governmentalregulations such as the Sarbanes-Oxley Act3, HIPPA4, and theGramm-Leach-Bliley Act5 making real-time information access anincreasingly competitive differentiator for many businesses.NETWORKED REMOTE STORAGE HAS THREE MAJOR TECHNICALREQUIREMENTS/CHALLENGES: A large volume of data must be transferred between sites. The transfer of that data is time-sensitive. Storage Input/Output (I/O) protocols require secondary processing in order to support high throughputs on distance.The first two points are intuitive, but the third is not. Storage protocols are handshakingprotocols designed for operation within a singledata center. When equipment gets remoted roughly in excess of 500 kms storage protocolstend to need channel extension technology tohelp keep the pace of I/Os high, and minimize theAdaptive WAN: Enabling Networked Remote Storage and Grid ComputingFigure 2. Long Distance Storage Networks3This act requires management to make a written assertion stating their responsibility for establishing and maintaining an adequate control structure and procedures for financial reporting.4The Health Insurance Portability and Accountability Act (HIPAA) requires companies in the health careindustry to provide administrative simplification, security, and privacy.5This act requires companies to give consumers privacy notices that explain the institution s information-sharing practices and to give consumers the right to limit some of the sharing of its information.The impact of Web Services on WANs is three-fold: Web Services will trigger an exponential growth in inter-sitetraffic volume. Low latency will be an increasingly critical networkcharacteristic. High priority packet identification will be very difficult.Untitled Documenteffects of latency. Latency has a doubleimpact it affects the responsiveness ofboth the LAN and storage-basedapplications.Since such long distance connections arealways leased, recurring costs must beminimized and bandwidth utilizationmaximized. This is where technologies likeprotocol-independent compression, anddynamic protocol load balancing comes intoplay. They guarantee that the leasedresource is always fully utilized, and themaximum amount of data can be squeezedthrough the long distance connection.Due to storage replication traffic sensitivity tolatency, and IP networking s difficulty with performance,many organizations have separate networks for their storageand LAN traffic, as depicted in Figure 3. In this example, three datacenters are interconnected with a WAN packet network. Servers existat all three sites and storage systems exist at two of the sites. Themain site has a protected connection to an Internet service provider.Leased lines interconnect the routers for LAN traffic. Separate leasedlines interconnect the storage systems. Additionally, there aremultiple network domains and repeated IT infrastructures.This is a very costly approach. Network performance is bounded bythe bandwidth prohibiting high-performance applications andcomplicating the rollout of new applications. The recurring costs ofmultiple leased services are high, and bandwidth cannot bedynamically shared between applications. If the storageadministrator must execute a restore from remote storage, thestorage processes cannot use bandwidth automatically from theother isolated WAN links to minimize the time it takes to restore. Inshort, the WAN has become an inflexible, scarce resource that isinefficiently used and expensive to maintain.It would be much more efficient if the network was flattened combining LAN and storage traffic on the same bandwidth. Ratherthan multiple domains, the result would be a much more manageablesingle domain, as shown in Figure 4.Architecturally, flattening the network seems very attractive, but does it cost less? First, router WAN blades, which are particularlyexpensive, can be eliminated. The recurring costs of repeated ITinfrastructures consolidating some servers and their associatedsoftware licenses are also eliminated. [Keep in mind that manymiddleware vendors list price their software according to physicalCPU counts. For example, eliminating a 10-15,000 two-way servercan also eliminate 50,000 of database software licenses ( 25,000per CPU).] And perhaps most importantly, flattening the network canreduce and perhaps eliminate the costs of specialized softwaredevelopment earlier referred to as tuning .4Adaptive WAN: Enabling Networked Remote Storage and Grid ComputingT1/T3 Linksfor LAN ExtensionT1/T3 Linksfor Storage ReplicationISPIPIPIPIPIPIPInternetIPDNS/DHCPServersFirewallServersStorageStorageServersFigure 3. Typical Inter-Site ConnectivityUntitled Document5Grid ComputingGrid Computing has gotten a lot ofattention recently. It seems every majorserver, storage and middleware vendorhas a grid initiative. Maybe you ve seen an IBM On-Demand commercial. HP is pushing Adaptive Enterprise .SAP has NetWeaver . Sun has its N1Architecture . Oracle has its 10g family.VeritasTMjust calls it what it is, GridComputing .No matter how you label it, GridComputing is being driven by the need tointeroperate between heterogeneouscomputer and storage assets. Rather thancreating silos of computational andstorage assets dedicated to a singlebusiness process, Grid Computingassets are pooled amongst allprocesses. This promises organizationsbetter asset utilizations and the ability tosize systems for the likely case. Whennetworks no longer have to be designedfor the worst case, significant cost savingsresult. For these reasons, many arepredicting a 20 billion market for GridComputing by 2010.Right now though, Grid Computing is in itsearly adopter phase. Scientists and engineersare using grids of computing assets spreadacross wide geographies for some heavycomputational calculations. They re saving money andcompleting their calculations faster than before, but their gridapplications are not what one would call time-sensitive.Adaptive WAN: Enabling Networked Remote Storage and Grid ComputingPrivateFiberISPIPIPIPInternetIPFirewallServersStorageStorageServersCienaCienaCienaFigure 4. Combined LAN and Storage ConnectivityDiskArrayDiskArrayDiskArrayDiskArrayDiskArrayServersServersServersTransactionTransactionTransactionSite ASite BSite CServer PoolStorage PoolDiskArrayRemote Backup SiteFigure 5. Multi-Site for Time-Sensitive Applications.Untitled Document6Time-sensitive applications those in which an end user customer isawaiting the result have only been implemented on grids within asingle location. The true promise of Grid Computing is theemployment of storage and computational assets throughout amultiple site organization. Today, using multiple site assets for atime-sensitive application requires too much code revision. Havingto recompile and debug applications so they are grid-enabled isslowing the grid s adoption of Grid Computing.Eventually, these Grid Computing architectural issues will beresolved. And grids will employ storage and compute assets spreadacross multiple sites for time-sensitive applications. When thathappens, organizations will need what only an Adaptive WAN cangive a network that provides high bandwidth, low latency andapplications transparency.The Importance of an Adaptive WANAs mentioned, Web Services, Networked Remote Storage and GridComputing all have significant impact on the inter-site/WAN trafficdemands of the enterprise. Exponential traffic growth, delaysensitivity, high priority packet identification and the support ofhandshaking storage protocols require a new type of networkingapproach that not only complements today s WAN, butaccommodates tomorrow s requirements.What are these requirements? WANs must be resilient to satisfythese changing needs. They must be scalable as inter-site trafficexpands. And they must be flexible to protect network investmentsas needs evolve. Requirements must be satisfied economically.What s needed is an Adaptive WAN leveraging Layer 1 and Layer 2transport with next generation flexibility technologies.Adaptive WANs provide enterprises with the ability to support anymix of applications between sites by providing all traffic with highperformance, zero packet loss and low fixed latency, regardless ofwhere the data is sourced from a remote LAN-based server, a SAN-based storage array or a legacy mainframe. An Adaptive WAN doesnot attempt to prioritize or accelerate one form of traffic at theexpense of another. It advocates a simpler approach by leveraginghighly reliable intelligent switched transport-based networks toprovide the throughput, the deterministic response times, and thehigh availability mission-critical applications require.Adaptive WANs are transparent to applications and allow flexibilityto support changing application requirements. Intelligent Layer 1and Layer 2 networks allow for application transparency due to theirability to provide local LAN-like connectivity across the WAN. UnlikeLayer 3, Layer 5 or even Layer 7 networks, they provide not only highbandwidth, but high application availability, near zero packet lossand low determinist latencies no matter what the load or applicationAdaptive WAN: Enabling Networked Remote Storage and Grid ComputingSite 1Site nSite 2Figure 6. Any application mix from any data source: LAN, SAN or MainframeUntitled DocumentAdaptive WAN: Enabling Networked Remote Storage and Grid Computing7mix. Not only does this allow traditional wide area client/serverperformance to be optimized, it also allows for new dynamiccomputing models including the wholesale adoption of WebServices and SOAs that can be built to run across multiple sites. Itlays the ground work for Grid Computing.WHAT ARE THE KEY REQUIREMENTS OR CHARACTERISTICS OFAN ADAPTIVE WAN? High throughput in the gigabit plus range Supports both LAN and storage traffic in any mix without impact Provides near zero packet loss to 100% line utilizations, even if load is continuous Provides low, predictable end-to-end application latency to matter what the application mix or load Simple to design, deploy, provision circuits, and monitor performance at the application protocol level Provides capabilities to detect and isolate faults and correct before they impact application performance Applications-Oriented Networking vs. Adaptive WANThere are two fundamentally different architectures being proposedto solve the emerging needs of Web Services, Networked RemoteStorage and Grid Computing. The first, which is sometimes referredto as applications-oriented networking, attacks the high prioritypacket identification challenge of Web Services. It looks inside thecontents of packets and makes switching decisions based on what itfinds. The second, Ciena s Adaptive WAN, focuses on the trafficvolume challenge which also alleviates the need for prioritization.Whereas applications-oriented networking switches operate at ISOLayers 5 and above, Adaptive WAN switches operate at Layers 1 and 2 where historically it has always been more efficient.The approach of intrusively inspecting each packet is problematic. Ateach switching point of an applications-oriented network, no matterhow efficient it is, one would have to execute a set of instructionsand transfer bits into and out of memory adding delay and latency. There are also some serious security issues. If network switches canlook inside a packet s contents, they could become fertile ground forhackers. Organizations have collectively spent billions trying to maketheir servers and storage systems hacker-proof. Should we assumethat they want to worry about their networking equipment too?Applications-oriented networking proponents believe that they canget by with less WAN bandwidth. But that s an illusion. Duringperiods of peak load, high priority packets are transmitted at theexpense of lower priority packets. Those lower priority packets are lost triggering packet retransmissions, which in turn also can be bounced, leading to even more packet retransmissions.Applications-oriented networking users will keep adding bandwidthuntil these packet storms subside. So, applications-orientednetworks will eventually converge on a similar amount of bandwidthas an Adaptive WAN. But these networks will go through a lot ofgrowing pain getting there!Adaptive WAN techniques are the superior approach. Unlikeapplications-oriented networking, Adaptive WANs satisfy the lowlatency requirements of Networked Remote Storage and multiplesite Grid Computing. It can be used for both storage protocols andTCP/IP. Adaptive WAN is the method being used today byenterprises for their most demanding application: remote storageconnectivity. With storage, lots of traffic must be sent with aminimum of delay. Storage networks switch traffic at the lowestpossible layer and maintain transparency to the application. Switching Prioritization Processing Performance Impact on Applications Security Risk Bandwidth RetransmissionsLayers 4 7Layers 1 2Packet ContentsHeader InformationSignificantMinimalMillisecondsMicrosecondsSignificant TuningTransparentHighCreeping Capacity IncreasesModerateLowWDM where available; DS-3, GbE or better (withL1/L2 optimization) elsewhereNegligibleApplications-Oriented NetworkingAdaptive WANTable 2. Competing ApproachesUntitled DocumentAdaptive WAN: Enabling Networked Remote Storage and Grid Computing8Applications transparency is a major advantage for Adaptive WANs.It allows software to execute as designed without tuning fornetwork constraints. Given that software developers typicallycomprise almost half an organization s IT budget and the networkportion is only around 10 to 15%, it could be more cost-effective to invest in bandwidth rather than applications tuning . Iforganizations take this holistic approach, they might may find thatnetwork investments can significantly lower their software develop-ment costs and hasten the time-to-market for new applications.Ciena s Adaptive WAN Approach Ciena s Adaptive WAN provides enterprises a simpler, lower costmeans to reliably extend their current and developing mission-criticalapplications between sites over both copper and optical lines.Ciena s approach provides gigabits of bandwidth with zero packetloss and low deterministic application latencies. This allows any mixof applications that run over LANs or SANs to be extendedtransparently across a consolidated multi-site WAN.Ciena s Adaptive WAN approach works without the needto convert all storage traffic to IP . ComplicatedQoS prioritization schemes are also avoided bydelivering all traffic with high quality anddeterministic application response times. Amix of intelligent switching of both LAN andstorage traffic scaling to 10s of gigabits isutilized, fed by DS3 access networks capableof supporting sustained real-time gigabit trafficthroughput. This all happens at or below Layer 2 in theprotocol stack without the need for continuous and costlyapplication tuning as advocated by application aware approaches.CIENA S UNIQUE SOLUTION CAPABILITIES INCLUDE: Support for both switched LAN and storage traffic in any mix without impact Deployment over copper and fiber based networks Support of maximum distance for synchronized mirrored storage between data centers as well as BC/DR Near zero packet loss to 100% line utilizations, even if load is continuous Low, predictable end-to-end application latency no matter what the application mix or load without complex application tuning schemes that must constantly be readjusted. Provides capabilities to detect and isolate faults and correct before they impact the applicationCiena s Adaptive WAN solution provides the simplest means to rollout of distributed computing architectures across multiple sites,while reducing new application time-to-market and total cost ofownership. Ciena s approach is not only more cost-efficient for ITdepartments to roll out, it is more effective in supporting theproductivity needs of the end business users.A variety of situations are solved with Ciena s Adaptive WANsolutions. In Figure 7 an example of a simple point-to-pointconnectivity between data centers. KEY ATTRIBUTES OF THIS SCENARIO INCLUDE: Low cost access Sites with or without fiber Zero packet loss Low deterministic latencyGbEFCESCONDS3 OCNCopper/Fiber AccessFigure 7. Simple Point-to-Point Adaptive WANUntitled DocumentThis approach utilizes the CN 2000TMStorage and LAN ExtensionPlatform to interconnect data centers. The CN 2000 not onlysupports BC/DR requirements, but with its compression and WANoptimization features, the platform is capable of supporting mirroredsynchronized SAN and high speed LAN extension over copperaccess links as well as over fiber links. The CN 2000 can generate a continuous throughput of up to 1 Gb/s over DS3 copper accesslinks all with zero packet loss and low deterministic response time.This approach can handle mainframe, SAN and LAN extensionsimultaneously over the same fibered link without applicationinterference and makes it the industry s most cost-effective andflexible data center interconnectivity solution.Extending the distance or adding other point-to-point sites maysometimes be required. A flexible extension of the point-to-pointapproach is shown in Figure 8. KEY ATTRIBUTES INCLUDE: Low cost access-sites without fiber Zero packet loss Low deterministic latency Distance can be extended with standard SONET/SDH hand-off.Figure 8 illustrates using the CN 2000 for data center access overDS3 copper or fiber access links. The CN 2000 is capable ofencapsulating the storage and LAN traffic in standard SONET/SDHpayloads for transport over any SONET or SDH network. Additionally, an organization can take advantage of the low cost CN 3606TMGFP-based SONET/SDH add/drop multiplexer (ADM) forcost-effective transport of the traffic over extended distances. TheCN 3606 provides low fixed delay and reliably delivers all trafficwithout loss. Unlike traditional SONET ADMs designed only forcentral office environments, the CN3606 is cost and sized optimizedto make it affordable for enterprise use.Figure 9 illustrates a scenario for a metropolitan area with severalsites needing switched connectivity and aggregation of multiple sitetraffic. This approach uses carrier-grade Ethernet switching with zeropacket loss. KEY ATTRIBUTES INCLUDE: No need to WAN mesh all sites Zero packet loss Ensures desired low deterministic end-to-end network latencyFigure 9 illustrates how low cost metro Ethernet transport can beused to flexibly connect multiple data centers together whileconsolidating storage and LAN protocols over the same low-costtransport. The CN 4350TMis the industry s first WAN based Ethernetswitch capable of handling any sustained traffic load up to 100%wire and/or switch utilization without dropping packets.And, it does this with low deterministic transit latencythrough the switch. This allows end-to-endapplication response time to be optimized and heldconstant. The CN 4350 can both switch andaggregate traffic on gigabit as well as 10 GigabitEthernet links. Its focus is to making it cost-effect toflexibly connect and scale inter-data centerconnectivity.The CN 2000 can be used to feed the CN 4350 with not onlyLAN, but also SAN and mainframe traffic by encapsulating it asstandard Ethernet frames. The CN 4350 provides the required zeropacket loss and low deterministic response time transport requiredto support mirroring storage traffic and mainframe connectivityacross the switched WAN.Adaptive WAN: Enabling Networked Remote Storage and Grid Computing9GbEFCESCONCopper AccessFigure 8. Flexible Point-to-Point Adaptive WAN Copper AccessUntitled DocumentTo scale the Adaptive WAN solution to support a greater number ofprotocols, ports and rates, we introduce the CN 4200TMFlexSelectTMAdvanced Services Platform. It provides maximum flexibility and canbe extended to any distance required. It is primarily intended fordata centers needing more ports and higher data traffic rates.The CN 4200 FlexiPort technology allows any protocol or rate to beprovisioned on any service port. This makes initial roll-out muchsimpler and less prone to error, while minimizing sparing and on-going costs. It is the only platform that can have ports re-provisionedas different protocols as well as rates, so it features the industry slowest total cost of ownership (TCO). It s easy toevolve from ESCON to FICONTMor FibreChannel (FC) as well as move up in capacity withjust a few key strokes from a managementconsole.The CN 4200 FlexSelect supports industrystandard generic framing procedure (GFP) toencode traffic onto wavelengths, as shown in Figure 10. The CN 4200 allows for simplehandoff at the wavelength-level for long-haultransport. It also allows for the add/drop ofindividual applications within a wavelengthwithout terminating the whole wavelength. Thismaintains high service quality while minimizing cost,as most competing solutions require dropping the entirewavelength and then retransponding the connections thatneed to be carried on to another destination.Ciena s Adaptive WANs are not just a theoreticalconcept. They re deployed in productionnetworks carrying mission-critical applicationstoday. Figure 11 depicts an Adaptive WAN inoperation at a Fortune 500 company. Within themetropolitan area, they re employing a DWDM.And their east cost data center 2100 miles away isinterconnected with a DS-3 circuit. The Cienasolution allows them to successfully implemented synchronous diskmirroring within the metropolitan area (up to spans of 300 miles) andasynchronous mirroring over 2100 miles all while carrying the inter-site LAN traffic. And, they re employing a mix of protocols: ESCON,Fibre Channel and TCP/IP .Connectivity diagrams such as Figure 11 are fairly common frommany vendors. It is important to note what is going on inside thoseconnectivity lines. At Ciena, we ve proven that we offer the lowestcost per moved bit. We are a leader in storage extension technology,which is the foundation of high performance Adaptive WANs. Cienaalso offers the best performance over any distance, whether across Adaptive WAN: Enabling Networked Remote Storage and Grid Computing10GbEGbEGbEFFGbEFESCONESCON10GbE WANOptical Access1000 TxFigure 9. Virtual Mesh Adaptive WANFICONESCONFC4G FC 10G FCGbENAS VolumesXML applicationsEthernet for VideoInter-office LANTape on ESCON ChannelFibre block I/OPrimaryMetroOTU2 Handoff of all SONET/SDH traffic including Ethernet-overPA-LM aggregation and switching of GF -F mapped EthernetLambda serviceswrapped in OTU2Inter regional site traffic handoffMetro site data center-1st and 3rd party hostGrooming of traffic from Access rings onto traffic-specific OTU1/2 Metro or Regional wavelengths4G FC10G FCGbEFigure 10. Flexible Scaling Adaptive WANUntitled Documentthe city, across the country or around the world. We address the fullbreadth of storage protocols over both dark fiber and serviceproviders SONET, SDH and Gigabit Ethernet infrastructures. And, wecarry the highest level of certifications from all major storage vendors.In this example, storage vendor certifications were very important.Many IT departments must be confident that networking productsinteroperate correctly with their multiple million dollar storagesystems. Ciena s proven interoperability and partnerships give ourcustomers that confidence. SummarySOA and the coming proliferation of Web Services will significantlyaffect WAN. Organizations will experience exponential traffic growthand will need WANs that provide high throughput with minimallatency. These needs are very similar to today s requirements ofNetworked Remote Storage. Organizations should adopt a WANstrategy similar to how they re solving storage system connectivitytoday what Ciena calls an Adaptive WAN. By executing anAdaptive WAN strategy, organizations can simultaneously solve threemajor network challenges: today s Networked Remote Storage, thenear term proliferation of Web Services; and future multiple site Grid Computing.Adaptive WANs are critical for multiple site organizations with time-sensitive applications. They enable organizations to place their ITAdaptive WAN: Enabling Networked Remote Storage and Grid Computing11Hosts AS/400, HP , Sun , WintelHosts: AS/400, HP, Sun, WintelOther LAN trafficOther LANtrafficOther LANtrafficESCONESCONESCONESCON/FICONFCFCMainframeMainframeDWDMDS3 Circuit:2100 miles20 milesData Centeron East CoastSite ASite BMidwestern CityCienaCienaCienaCienaFigure 11. Adaptive WAN in ProductionUntitled Documentassets anywhere including their storage assets. Unlike applications-oriented networking, Adaptive WANs handle both TCP/IP andstorage protocols, and minimize applications tuning a budgetbuster for most IT organizations. With a highly reliable network with high bandwidth, deterministic end-user response and physicalLayer security, you can deploy new applications faster a clear Layersecurity, you can deploy new applications faster a clear competitiveadvantage.Ciena s Adaptive WAN solution is field-proven in the mostdemanding production networks today. Carrying mission-criticaltraffic with a full variety of bandwidth services (e.g., SONET/SDH,DS-3, GbE, DWDM). Ciena s Adaptive WANs carry the highest levelof certifications from all major storage vendors.Adaptive WAN: Enabling Networked Remote Storage and Grid Computing12Ciena may from time to time make changes to the products or specifications contained herein without notice. All trademarks andregistered trademarks are the property of their respective companies. 2005 Ciena Corporation. All rights reserved. WP028 5.2005Ciena Corporation is the network specialist, focused on expanding the possibilities for its customers' networks while reducing their costs of ownership. The company's systems, software and services target and cure specific network pain points so that telcos, cable operators, governments and enterprisescan best exploit the new applications that help them achieve their business and economic goals. For more information please call 1-800-207-3714 (US) or1-410-865-8671 (outside US) or visit www.ciena.com.