Layer 3 Switching: An Introduction by Rob Ciampa


Published on

What was initially a marketing whitepaper on technology, this document became part of graduate theses, channel strategy, and technology marketing. When I wrote it, it was a way to reposition and redefine existing technology (routing) and the industry leader (Cisco), so that 3Com could penetrate the router market without appear to go head-to-head with the leader.

From the intro:

Many people involved in the deployment of information technology compare their professionto the world of Indiana Jones, a Hollywood action hero of great intelligence, challenged by friendly and unfriendly forces while searching the for lost cities and hidden treasure. Like Indiana Jones, IT managers might not have a chance if they make a wrong decision. In, the journey through the jungle full of deception, wrong turns, and stumbling.

In a competitive environment, the proper decisions can catapult corporations their competitors, barriers to entry redrawing the battlefield. A prime example, which has used the Internet to the bookselling industry, at the of formidable competitors who now to mimic the techniques of their upstart. On the other hand, blindly following temporary technology fashion may leave IT managers in the jungle, or out of a job. paper is a map through the jungle of infrastructure, particularly on how Layer 2 switching and Layer 3 have combined to form the powerful 3 switching architecture. The paper analyzes Layer 3 switching from both a functional an operational, helping the make an informed assessment of its merits an enabling technology.

, and how best to reposition incumbent players in changing marketing.

Published in: Technology
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Layer 3 Switching: An Introduction by Rob Ciampa

  1. 1. 3 C o m Te c h n i c a l P a p e r s ® Layer 3 SwitchingAn Introduction
  2. 2. Layer 3 SwitchingAn IntroductionContentsA Brief History of Protocol Layering 2Effective Information Management 2Layering 101 3Contemporary Layering Model 3 Layer 1 4 Layer 2 4 Layer 3 5 Layer 4 6 Layer 5 6Enter the Layer 3 Switch 6 A Layer 3 Switch Is a Router 7 Router Interfaces as Layer 2 Switching Domains 7 Effective Application of Policy 7 Ease of Management 8Layer 3 Switching v. Traditional Routing 8The Business Case for a Layer 3 Switch 8 Cost Savings 10Case Study 10 Initial Network 11 Phase 1 11 Phase 2 12 Phase 3 12Future Trends 12Conclusion 13 1
  3. 3. Acronyms and Layer 3 Switching architecture. SNA’s formal counterpart, theAbbreviations International Standards Organization (ISO) An IntroductionASIC Open Systems Interconnect (OSI) model,application-specific By Robert Ciampa which was a seven-element layout, succeededintegrated circuit more as a pedagogical tool than as an imple- Many people involved in the deployment of mentation foundation. As a result, manyATMAsynchronous Transfer Mode information technology compare their profession academics, along with the some rare imple- to the world of Indiana Jones, a Hollywood menters (Digital Equipment Corporation withBGP DECnet Phase V) were left in the networkingBorder Gateway Protocol action hero of great intelligence, challenged by friendly and unfriendly forces while searching the jungle.CoS world for lost cities and hidden treasure. Like Meanwhile, the Internet Protocol (IP)Class of Service Indiana Jones, IT managers might not have a was enjoying some deployment success, first second chance if they make a wrong decision. In through the U.S. Department of Defense’sDHCPDynamic Host Configuration fact, the journey through the networking jungle ARPANET—the genesis of the Internet—Protocol is full of deception, wrong turns, and stumbling then into diverse university communities. IP blocks. and its associated higher-layer protocols, suchDNS In a competitive environment, the proper as User Datagram Protocol (UDP) and Trans-Domain Name System technology decisions can catapult corporations port Control Protocol (TCP), were supposedFDDI to be supplanted by the OSI protocols, but over their competitors, erasing barriers to entryFiber Distributed Data the increasing complexity of OSI, exacerbatedInterface and redrawing the battlefield. A prime example is, which has used the Internet to by a prolonged ratification process, underminedFIRE revolutionize the bookselling industry, at the its prospects. IP continued to be deployed,Flexible Intelligent Routing while other IP-like protocols such as NetWare’s expense of formidable competitors who nowEngine IPX and Apple’s AppleTalk were enjoying their attempt to mimic the techniques of their upstartIEEE foe. On the other hand, blindly following tempo- own success. The similarity among IP, IPX,Institute of Electrical and rary technology fashion may leave IT managers and AppleTalk is no accident: they share aElectronics Engineers common lineage through Xerox Network Sys- stuck in the jungle, or out of a job.IETF This paper is a map through the jungle of tems (XNS), an older but simpler model thanInternet Engineering Task internetworking infrastructure, particularly OSI.Force focusing on how Layer 2 switching and Layer 3 routing have combined to form the powerful Effective Information ManagementIPInternet Protocol Layer 3 switching architecture. The paper ana- Just as there are many types of jungles, so lyzes Layer 3 switching from both a functional there are many types of data networks. AndIPsec jungles and networks have some striking simi-IP Security and an operational perspective, helping the reader make an informed assessment of its merits larities in the way they are organized. In theISO as an enabling technology. jungle, the parts of the whole are calledInternational Standards ecosystems; in the network, they are called lay-Organization ers. Each subsystem, or layer, is often quite A Brief History of Protocol LayeringISP To fully appreciate Layer 3 switching, it is use- distinct from others within the same system orInternet service provider ful to examine its ancestry, since many com- network, but depends upon access to the oth- mon traits still prevail. Rather than go back to ers for its survival. Call it the food chain orMAC the stone age of hierarchical networking, we’ll call it the protocol stack.Media Access Control begin with the “modern era” of data commu- Knowledge of layering is crucial for theMIB strategic and tactical deployment of both net- nications, a time of peer-to-peer networkingManagement Information working and information technology in anBase with heterogeneous systems. It is interesting to note that hierarchical networking—its best organization. Many people view layering as anMPOA example being IBM’s Systems Network academic exercise in which Layer 2 representsMultiprotocol over ATM switching and Layer 3 represents routing. Architecture (SNA)—was probably the quin- tessential—but immutable—client/server Such shortsighted thinking leaves many orga- nizations at the mercy of the performance 2
  4. 4. constraints of their collapsed backbone Application Acronyms androuters. Understanding the capabilities and Abbreviations Presentationlimits of each layer is the foundation for infor- NHRPmation management. Strategic decisions must Session Next-Hop Resolutionbe made about application deployment, net- Transport Protocolwork scalability, performance, and cost of Network OSIownership. Tactical decisions must be made Data link Open Systems Interconnectabout which products to apply as part of an Physicaloverall solution. This methodology becomes OSPF Open Shortest Path Firsteven more important as voice, video, and data Figure 2. OSI Layering Modelnetworks continue to converge, blurring the QoSonce clear demarcation between data commu- Quality of Servicenications and telecommunications. other stacks and entirely proprietary, but com- RIP munication between peers in different stacks Routing InformationLayering 101 must be open and consistent. The notion of ProtocolAlthough this paper is about Layer 3 switch- open systems has been a major factor in the RMONing, a quick overview of layering is needed. growth and operation of the Internet, along Remote MonitoringLayering schemes provide guidelines, rather with those of institutional organizations. It isthan strict rules, for delegating networking also important to note that an element at a RSVPfunctionality. Figure 1 shows the basic princi- particular layer may be further broken down Resource Reservation Protocolples of layering. Elements at the same layer, into additional layers. This is most clearly seenshown on the horizontal, are known as peers with Asynchronous Transfer Mode (ATM) RTPand communicate via a well-known (and doc- models. Finally, in certain models, higher lay- Real Time Protocolumented) protocol. Messages are exchanged ers may share information with lower layers to SDHamong peers, the protocol defining the for- either conserve system resources or improve Synchronous Digitalmat, syntax, semantics, and sequencing. Ele- performance. The Internet Engineering Task Hierarchyments within the same stack, shown on the Force (IETF) Next-Hop Resolution Protocol SLAvertical, communicate via an internal interface. (NHRP) is an example of this intra-layer com- Service Level AgreementThis interface, though usually not well docu- munication, allowing Layer 3 “shortcuts.”mented nor standard, often exhibits the same This concept will be discussed later. SNA Systems Networkcharacteristics as a protocol, the only differ- Architectureence being that the interface protocol between Contemporary Layering ModelLayer n and Layer n + 1 on stack 1 may be For many years, the OSI model (Figure 2) was SONETwholly different from that of stack 2. the reference layering paradigm for data net- Synchronous Optical Network As mentioned, communication within working. The OSI model was an extremelyone stack may be different from that within powerful architecture that included well- TCP defined Layer n ⁄ Layer n + 1 protocols in Transport Control Protocol addition to rich peer-to-peer protocols. Unfor- UDP tunately, much of this model succumbed to User Datagram Protocol the complexity of the protocols and the effects Layer n + 1 Layer n + 1 of an overly rigorous standardization process. VPN virtual private network Since only a few elements survived to become part of the contemporary networking model, WAN Layer n Layer n no further analysis will be made of this model. wide area network The contemporary network layer archi- WinSock tecture is much simpler than its OSI counter- Windows Sockets Layer n – 1 Layer n – 1 part. Originating from various research and XNS defense initiatives, the contemporary model Xerox Network Systems was intended to be supplanted by OSI.Figure 1. Layering Reference Model Instead, it became the de facto networking 3
  5. 5. Layer 2 Application This layer, known as the switching layer, Transport allows end station addressing and attachment. Routing Because architectures up to Layer 2 allow end Switching station connectivity, it is often practical to Interface construct a Layer 2–only network, providing simple, inexpensive, high-performance con- Figure 3. Contemporary Layering Model nectivity for hundreds or even thousands of end stations. The past five years have seen the extraordinary success of the “flat” network standard, especially through IP. As mentioned, topologies provided by Layer 2 switches con- both IPX and AppleTalk are quite similar to nected to other Layer 2 switches or ATM IP, but are slowly becoming less prominent as switches. IP dominance continues to grow. This discus- Layer 2 switching, also called bridging, sion will emphasize IP, but the methods dis- forwards packets based on the unique Media cussed can easily be applied to environments Access Control (MAC) address of each end using NetWare and Apple protocols. station. Data packets consist of both infra- Figure 3 shows the contemporary net- structure content, such as MAC addresses and working model based upon IP. Network par- other information, and end-user content. At ticipants, whether infrastructure equipment Layer 2, generally no modification is required (switches and routers) or end systems (clients to packet infrastructure content when going and servers), may include some or all of the between like Layer 1 interfaces, like Ethernet protocol stack. to Fast Ethernet. However, minor changes to infrastructure content—not end-user data Layer 1 content—may occur when bridging between This layer, known as the interface layer, is unlike types such as FDDI and Ethernet. responsible for device connectivity. Though Either way, processing impact is minimal and usually represented by well-known network so is configuration complexity. types—Ethernet, Fast Ethernet, Gigabit Eth- Layer 2 deployment has seen the most ernet, Token Ring, FDDI, ATM, SONET/ striking infrastructure change over the past SDH, etc.—Layer 1 also covers the subtypes. decade. Shared Ethernet, represented by par- For example, Fast Ethernet provides physical ticular cable types or contained within shared connectivity over copper media (100BASE- hubs, offered a very simple, and even more TX) and over fiber media (100BASE-FX). inexpensive, approach for Layer 2. Though Fiber can be further divided into multimode still quite popular, shared technology, where or single mode, with single mode further par- all stations use the same bandwidth slice, has titioned based on its “reach,” the distance over very limited scaling capabilities. Depending which it can transmit. Certain technologies upon the applications being used, shared net- are actually used as a pure Layer 1 element works of more than one hundred users are (SONET/SDH) or provide a virtual Layer 1 becoming less common. Many network element (ATM with SONET/SDH). designers have “tiered” their infrastructure by While the various types of Ethernet are feeding shared Layer 2 into switched Layer 2 rather straightforward, FDDI, ATM, and or even Layer 3. Switched Layer 3 apportions SONET/SDH add more complexity, while each station—or port—its own dedicated providing extended Layer 1 capabilities such bandwidth segment. Recent enhancements at as fault tolerance and support for physical Layer 2 provide packet prioritization capabili- multiplexing to support distinct traffic flows ties for the application of network policies. such as voice and data. With these added The new IEEE 802.1p standard defines Class capabilities comes added cost, and sometimes of Service (CoS) policies capabilities for Layer slower performance. 2 segments.4
  6. 6. Note that Layer 2 does not ordinarily routers maintained a presence at many pointsextend beyond the corporate boundary. To within a corporate network. For a while thisconnect to the Internet usually requires a presented minimal problems, since a majorityrouter; in other words, scaling a Layer 2 net- of the data traffic stayed local to the subnet,work requires Layer 3 capabilities. which was increasingly being serviced by a Layer 2 switch. But concurrent with theLayer 3 increasing acceptance of Layer 2 switching asThis layer, known as the routing layer, pro- an essential component of network infrastruc-vides logical partitioning of subnetworks, scal- ture were two other developments: the migra-ability, security, and Quality of Service (QoS). tion of servers to server farms for increasedQoS, a recent enhancement to Layer 3, goes security and management of data resources;beyond the simple packet prioritization found and the deployment of intranets, organization-in CoS by providing bandwidth reservation wide client/server communications based onand packet delay bounding. Web technology. These factors began moving The backbone of the Internet, along with data flows off local subnets and onto thethose of many large organizations, is built routed network, where the limitations ofupon a Layer 3 foundation. IP is the premier router performance increasingly led to bottle-Layer 3 protocol. In addition to Layer 2 MAC necks.addresses, each IP packet also contains source With the routers causing informationand destination IP addresses. For an intranet flow constriction, IT managers becamepacket, one IP address addresses the client, the increasingly reluctant to deploy new, enablingother the server. technologies, such as multicast-based applica- IP in itself is not a particularly complex tions and middleware. Even the migration ofprotocol; extensive capabilities are supplied by desktops to higher-performance media con-the other components of the IP suite. The nections, such as 100 Mbps Fast Ethernet,Domain Name System (DNS) removes the were scrutinized as long as 10 Mbps routerburden of remembering IP addresses by asso- interface funnels were in place.ciating them with real names. The Dynamic Router vendors attempted to respond byHost Configuration Protocol (DHCP) eases offering higher-performance interface cards,the administration of IP addresses and is used but throughput was fundamentally boundedextensively by network administrators and by centralized, software-based architecturesInternet service providers (ISPs). Routing pro- that simply could not go any faster. The sametocols such as Open Shortest Path First software responsible for managing WAN links,(OSPF), Routing Information Protocol (RIP), X.25, and asynchronous terminal lines wasand Border Gateway Protocol (BGP) provide now expected to handle next-generation giga-information for Layer 3 devices to direct data bit networks. Router vendors tried distribut-traffic to the intended destination. IP Security ing functionality to improve performance,(IPsec) furnishes elements necessary for secu- resulting in a hodgepodge collection of routerity, such as authentication and encryption. IP processing and interface cards. Was the devicenot only allows for user-to-user communica- still routing, or was it performing some othertion, but also for efficient dissemination over packet forwarding scheme?point-to-multipoint flows, known as IP Multi- Emerging QoS was even more suspect.cast. Higher-layer protocols, discussed later in The IETF was moving forward on Resourcethis paper, provide even greater versatility for Reservation Protocol (RSVP), a signalingcontent distribution. method to set up bandwidth and delay control Although many organizations received on packet-based internetworks. Monitoringtremendous performance advantages by con- RSVP flows, using a process know as policing,verting routed and shared networks to Layer 2 required extensive software support on alreadyswitching, it became apparent that some level overburdened legacy routers. Could QoS beof partitioning was still required. Consequently, practical on a contemporary LAN? 5
  7. 7. Meanwhile, standards bodies such as the rather, they are written to interface with a spe- ATM Forum were working on methods to cific communication library, like the popular offload the Layer 3 bottleneck by exploiting WinSock library available in Windows-based the capabilities of the lower layers. One result workstations. was the Multiprotocol over ATM (MPOA) When defining the behavior of the appli- specification, which uses Layer 3 routing cations they are writing, developers decide on information and the IETF’s NHRP protocol the type of transport mechanism necessary. to offload the routers and provide forwarding For example, database or Web access requires at the physical (ATM) layer. A Layer 3 switch robust, error-free access and would demand can route at Layer 3 or utilize MPOA; the per- TCP, though it could be implemented with formance is identical. more code and in a more cumbersome man- ner with UDP. Multimedia, on the other Layer 4 hand, cannot tolerate the overhead of connec- This layer, known as the transport layer, is the tion-oriented traffic and will commonly make communication path between user applica- use of UDP. For prioritization, either TCP nor tions and the network infrastructure and UDP can be selected, depending on the appli- defines the method of communicating. TCP cation or other parameters such as time of day. and UDP are well-known examples of ele- Any assistance that a network device can pro- ments at the transport layer. TCP is a “con- vide in terms of prioritization of the applica- nection-oriented” protocol, requiring the tion would be extremely beneficial to the establishment of parameters for transmission network manager, particularly during times of prior to the exchange of data. Web technology traffic volume from the LAN to the WAN. is based on TCP. UDP is “connectionless” and requires no connection setup, which is espe- Enter the Layer 3 Switch cially important for multicast flows. Elements Traditional routers, once the core components at this level also differ in the amount of error of enterprise networks, became a major obsta- recovery provided and whether or not it is vis- cle to the migration toward next-generation ible to the user application. Both TCP and networks. All the magic and alchemy involved UDP are layered on IP, which has minimal in trying to make a software-based router for- error recovery and detection mechanisms, ward packets more quickly created only an leaving the burden at Layer 4 or higher. TCP illusion of serious Layer 3 routing performance. forces retransmission of data that was lost by A fundamental shift in technology was the lower layers, whereas UDP makes the required. application responsible. In 1992, 3Com, a pioneer in both Layer A major enhancement to multimedia sup- 2 switching and traditional routing, began port at Layer 4 is the Real Time Protocol integrating its switching and routing products. (RTP). RTP works in conjunction with UDP, The motivation was twofold: to reduce the and provides services necessary for packet tim- number of devices to be managed, and to ing and sequencing. Many time-sensitive lower the cost of a combined Layer 2 and 3 applications running over IP networks now solution. Though the first solution was mostly actually include both UDP and RTP. software based, subsequent products displayed increasing use of application specific integrated Layer 5 circuits (ASICs)—first for address table queries, This layer, known as the application layer, then for forwarding packets. In 1997, 3Com provides access to either the end user or some delivered its third-generation, fully ASIC- type of information repository such as a data- enabled Layer 3 CoreBuilder™ 3500 switch, base or data warehouse. Users communicate based on the Flexible Intelligent Routing with the application, which in turn delivers Engine (FIRE) ASIC. For more information data to the transport layer. Applications do on FIRE, see the white paper “Flexible Intelli- not usually communicate with the lower layers; gent Routing Engine (FIRE): The Third-6
  8. 8. Table 1. Layer 3 Switching Product Evolution Generation Technology Product Routing Performance First Software LANplex® 5000 switch 50K pps Second ISE ASIC CoreBuilder 2500, 6000 switch 100K–1.1M pps Third FIRE ASIC CoreBuilder 3500, 9000 switch 3.5M–64M ppsGeneration Layer 3 Switching Architecture together, whether based on physical character-from 3Com,” available at istics or protocol information, is an extremely Table 1 shows 3Com’s Layer 3 switching powerful tool for network designers concernedproduct evolution. with capacity planning. This architecture is inherently scalable, capable of supportingA Layer 3 Switch Is a Router numerous external Layer 2 switches that resideVendors and the trade press alike have tried to either in the data center or the wiring closet.apply the term “Layer 3 switch” to various Such a design model preserves the subnet-products of the day, succeeding only in con- ted infrastructure, concurrently boosting per-fusing IT decision makers. This paper aims to formance of those subnets and enabling theremove that confusion. A Layer 3 switch does deployment of switched 10, 100, or 1000everything to a packet that a traditional router Mbps right to the desktop if so desired. Thedoes: concept of “subnet preservation” is the key to• Determines forwarding path based on effective and trouble-free network migra- Layer 3 information tion—it allows gradual migration, helping IT• Validates the integrity of the Layer 3 header managers to work within their staffing con- via checksum straints without the need to renumber and• Verifies packet expiration and updates reassign their entire network. accordingly• Processes and responds to any option infor- Effective Application of Policy mation As previously stated, contemporary Layer 3• Updates forwarding statistics in the Man- switches perform their forwarding—whether agement Information Base (MIB) Layer 2, Layer 3, unicast, multicast, or broad-• Applies security controls if required cast—in hardware. Software is deployed to Because it is designed to handle high-per- handle network administration, table manage-formance LAN traffic, a Layer 3 switch can be ment, and exception conditions. Some tech-placed anywhere within a network core or nologists view the hardware component of abackbone, easily and cost-effectively replacing Layer 3 switch as inflexible. In fact, hardwarethe traditional collapsed backbone router. The provides the ultimate flexibility not only inLayer 3 switch communicates with the WAN performance, but in parallel processing asrouter using industry-standard routing proto- well. The parallel processing model allows thecols like RIP and OSPF. network device to perform far more opera- tions on packets than previously imagined,Router Interfaces as Layer 2 Switching Domains especially with respect to the application ofThe Layer 3 switch has inherent Layer 2 policy.switching domains per interface, allowing for A policy is a mechanism to alter the nor-individual subnet bandwidth allocation, along mal forwarding of a packet through a net-with broadcast containment. Not all interfaces working device. Familiar examples includeare created equal, so the ability to group ports security, load balancing, and protocol option 7
  9. 9. processing. Newer policies include QoS, a way For the network management application to allocate bandwidth and control propagation perspective, a Layer 3 switch behaves exactly delay, in addition to CoS, a way to manage as a legacy router does. Because of its Layer 2 packet prioritization. QoS and CoS policies component, extensive Remote Monitoring are not only meant to enable new multimedia (RMON) capabilities are available. However, applications, such as LAN telephony, but to since Layer 3 and Layer 4 capabilities are pre- ensure network response time for mission- sent in the Layer 3 switch, higher-layer moni- critical applications, such as telemedicine. toring is available with RMON2 technology. Policy implemented by intelligent networking RMON and RMON2 have historically been devices, such as Layer 3 switches, enables the deployed with expensive external devices integration of voice, video, and data onto the known as probes. Moving the RMON/ same infrastructure, a process 3Com calls RMON2 capability into the Layer 3 switch is convergence. a major benefit for IT administrators. Software-based architectures cannot seam- lessly administer policy controls at even mod- Layer 3 Switching vs. Traditional Routing erate rates of speed (beyond 10 Mbps). The By now, it should be clear that a Layer 3 Layer 3 switch solves the problem, enabling switch can be deployed anywhere in the LAN policies to be applied at the same performance where a traditional router can be or has been levels as ordinary Layer 2 and 3 forwarding. used. Further innovation allows the Layer 3 switch Table 2 compares the two types of to apply policy based on Layer 4 information, devices. The Layer 3 switch has been opti- such as TCP and UDP port information. For- mized for high-performance LAN support and ward thinkers refer to this as “Layer 4 switch- is not meant to service wide area connections ing.” The FIRE architecture supports all these (although it could easily satisfy the require- policies, all the way to Layer 4. ments for high-performance MAN connectiv- Even with the massive capacity additions ity, such as SONET). This optimization boosts being planned for many networks, effective the performance of a Layer 3 switch to as much policy management enabled by Layer 3 as ten times that of a legacy router, while dri- switching is key to the protection and avail- ving the price down to as little as a tenth. This ability of critical resources. cost comparison does not include the lower training costs for Layer 3 switch administra- Ease of Management tors or the increased productivity of a high- One of the critical success factors for the Layer performance network. 2 switch was its implementation and opera- There is another major architectural dif- tional simplicity. Deployment was often as ference between a Layer 3 switch and a router. easy as powering on the switch, assigning it an A traditional router organizes bridging (Layer IP address, and making the physical network 2) and routing (Layer 3) as peers. A Layer 3 connections. Routers, on the other hand, switch layers routing on top of switching, per- required extensive training and forced users to mitting a more natural networking architec- sift through a multitude of arcane commands. ture while greatly facilitating scalability. Layer 3 switches remove such complexity. Setting up a routed environment is as simple The Business Case for a Layer 3 Switch as setting up a Layer 2 switch, defining the Some IT managers may be concerned about routed interface, and enabling the routing deploying a “new” technology such as Layer 3 protocols. IT managers concerned about their switching to their network. But Layer 3 investment in training staff on traditional switching is really an integration of two router platforms must assess whether this is proven technologies: switching and routing. truly an investment, or simply a sunk cost In fact, some Layer 3 switches are running the based upon vendor lock-in schemes. exact same routing software that has been fully tested and used in mission-critical networks8
  10. 10. Table 2. Layer 3 Switch vs. Legacy Router Characteristic Layer 3 Switch Legacy Router Routes core LAN protocols: IP, IPX, AppleTalk Yes Yes Subnet definition Layer 2 switch domain Port Forwarding architecture Hardware Software RMON support Yes No Price Low High Forwarding performance High Low Policy performance High Low WAN support No Yesfor nearly a decade. So whether the decision enables a variety of steps to reduce costs andmaker is an early adopter of technology or enhance security and business operations. Themore conservative, the Layer 3 switch can sat- following are examples of several such steps.isfy both needs. • Server farms. Today, the viability of many The first step toward the deployment of organizations is closely related to their intel-next-generation IT infrastructures is to ignore lectual property, often stored on databasesthe networking element. Although this may or server devices. The security and protec-seem absurd, it allows managers to focus on tion of these servers has been a major goalthe end users, services, and data without being of IT managers, who have been at odds withbound by historical network deficiencies. The the users of those servers. The point of con-network should be transparent. When the tention has been the dependence of serverrequirements for information transfer are performance on the network topology. Theknown, capacity planning techniques will response has been to move servers withindetermine the necessary client and server the same subnet or Layer 2 switchinginterconnects. Organizational and security domain as users. With data traffic patternsmandates are then applied, yielding the policy becoming more distributed, this approachand subnetted infrastructure. Cost is then was breaking down. The Layer 3 switchfactored in. Finally, the decision is made allows servers to be centralized with no per-regarding the appropriate networking prod- formance penalty, eliminating the cost ofucts to satisfy these requirements. numerous server repositories while keeping Layer 3 switching technology must be end users satisfied.considered from two perspectives. First, as a • Intranets. Because of its secure nature,migration tool to move users forward to along with its higher capacity, the intranet ishigher-performance networking, or surpris- becoming a viable corporate communica-ingly, to squeeze more performance out of tions vehicle with usage that includes HRwhat is currently installed. Many users com- record retrieval, major announcements,plain about FDDI performance, only to dis- computer-based training, and live videocover that the network is running at less than broadcasts. Delivering a wide variety of ser-20 percent of capacity. The problem is not the vices, some requiring a huge amount ofnetwork, but rather the devices attached to it. bandwidth, can wreak havoc on the oldThe second perspective addresses what can be router. The Layer 3 switch, because of itsdone when network performance bottlenecks higher performance, traffic prioritization,are removed. A high-performance network 9
  11. 11. and subnet preservation, is ideally suited for port. Surprisingly, greater densities can be the deployment of intranets. achieved with Layer 3 switching, freeing up • Converged networks. For some time, tech- valuable rack space and saving on physical nological prognosticators have been extol- cabling plant expansion. Training costs plum- ling the virtues of multimedia and warning met, too, as a four- to seven-day legacy router of the excessive demand it will place on IT course is replaced with a one- or two-day class infrastructures. But many network man- for the Layer 3 switch. agers have been disinclined to integrate their Major savings also lie in the ancillary voice, video, and data traffic, concerned not effects of applying Layer 3 switching technol- only with the bandwidth requirements, but ogy. Cost savings realized through server cen- fearing the degraded quality of the respec- tralization, notably in physical plant and tive elements. The ability to recognize and security, can be substantial, especially when respond to the unique attributes of voice, space is at a premium. Other, less tangible data, and video not only makes their inte- effects include improved response time and gration viable, but also attractive from a cost conformance with SLAs. Clearly, the overall and management perspective. The inherent cost of ownership benefits of Layer 3 switches flow recognition capabilities of Layer 3 versus routers can be substantial. switching enable practical deployment of converged networks without performance Case Study uncertainties. The following application scenario starts with a common contemporary network topology Cost Savings and illustrates a migration path toward a next- A traditional router may run U.S. $8,000 to generation infrastructure. The deployment $10,000 per Fast Ethernet interface, while a objectives are as follows: Layer 3 switch costs less than U.S. $1,000 per • To minimize network disruption net Inter WAN orate Corp cy Lega r outer uter cy ro ing Lega Serve r FDDI r r route Legacy cy b Lega ed hu uter Shar r ro Serve er Cent Data b ed hu Shar t Clien nt rtme Depa er Clien t serv nt rtme t Depa Clien Figure 4. Initial Network Configuration10
  12. 12. net Inter WAN orate Corp cy Lega r route uter cy ro rin g Lega Serve r FDDI tch r 3 swi Laye tch ed r 3 swi Shar Laye hub r Serve er Cent Data b ed hu Shar t Clien r Serve t nt Clien rtme Depa t ClienFigure 5. Data Center and Workgroup Upgrades to Layer 3 Switching• To preserve subnet infrastructure net for each protocol. Two of the routers ser-• To avoid parallel network construction vice WAN access: one for corporate network extension, the other for Internet service.Initial NetworkThe network core, shown in Figure 4, consists Phase 1of an FDDI backbone running at 20 percent The first phase (Figure 5) consists of keycapacity, occasionally peaking at 40 percent. legacy router replacement for the data centerCollapsed backbone routers are the connec- and for the most heavily used departments. Iftion points to the FDDI backbone, with the other legacy protocols such as DECnet orexception of some data center servers that Banyan VINES are present, the Layer 3 switchattach directly to the backbone. The legacy and the router can be co-located in the wiringrouters supply mostly 10 Mbps Ethernet closet or the data center, the Layer 3 switchinterfaces, with some 100 Mbps Fast Ethernet becoming the “express lane” for the contem-interfaces. Some of these Ethernet interfaces porary protocols. With the Layer 3 switch inconnect to Layer 2 switches, which then cas- place, the department and data infrastructurecade to hubs, while others connect to hubs behind it can then be upgraded to higher-per-directly. A majority of the desktops are shared formance Layer 2 switches, ultimately bring-10 Mbps Ethernet. Some of the servers are ing switched Ethernet to the desktop. Theswitched. Departments may have their own migration of key departmental servers toserver co-located on a subnet. The network is server farms may begin at this point. If therunning IP and IPX, with the subnets for both capacity required for the aggregate client-protocols aligned with the other. The FDDI server flows exceeds that of FDDI, the gradualring contains one subnet for each protocol, evolution of the backbone may begin at thisand each router interface also provides a sub- point, otherwise it will be covered in phase 2. 11
  13. 13. W AN orate Corp net Inter itch Serve r itch r 3 sw r 3 sw Laye Laye uter cy ro Fast Lega it Gigab rnet net itch Ether r 3 sw Ethe ring Laye FDDIServer er Cent Data itch r 3 sw Laye r2 Laye ch swit b ed hu Shar t Clien t nt Clien rtme Depa er Clien t serv nt rtme Depa Figure 6. Backbone Migration Begins Phase 2 access and signaling mechanisms for CoS and The second phase (Figure 6 on page 12) con- QoS. With the infrastructure distributed, yet tinues the replacement of the routers on the overlaid with a logical management structure, FDDI backbone with Layer 3 switches. The performance metrics can be tuned and modi- routers that were servicing the WAN connec- fied, giving greater viability to Service Level tions remain, but are now removed from the Agreements (SLAs). The legacy backbone has backbone and connected via Ethernet or Fast now been entirely eliminated and replaced by Ethernet to a Layer 3 switch. The migration higher-performance Gigabit Ethernet, ATM, of the backbone begins at this stage with the or both. The new backbone is inherently scal- choice of Gigabit Ethernet or ATM. (This able and is ready for any future network evo- choice depends upon a variety of factors, lution. Though beyond the scope of this which are beyond the scope of this paper.) paper, the core network will also become the The 3Com Layer 3 switching methodology is termination point for virtual private networks fundamentally unaffected by the choice (VPNs) as remote offices access the corporate between Gigabit Ethernet and ATM. In fact, infrastructure via the Internet. For more infor- the backbone could very well support both. mation on VPNs, see the white paper “Private Use of Public Networks,” available at Phase 3 The third phase (Figure 7) completes the evo- lution of the backbone, and introduces policy Future Trends services into the infrastructure. Such policy, The Layer 3 switching solution does not stop administered by the network manager, may here. Expect more Layer 4 capabilities to extend as far as the desktop, enabling network become available, handling advancements in 12
  14. 14. r serve net te a ccess Inter Remo WAN orate ATM VPN Corp Fast Fast net net Ether uter Ether cy ro Lega it Laye r3 Gigab rnet r h Ethe y serve switc r Polic itch Serve it r 3 sw Gigab t Laye Eth erne it itch Gigab rnet r 2 sw itch Ethe Laye r 3 sw it Gigab t Laye e Serve r E thern er Cent Data r3 Laye sw itch r2 Laye ch h swit r2 switc Laye t Clien Clien t nt rtme Depa er t ser v Clien nt rtme DepaFigure 7. Backbone Upgrade Complete; Policy Enabledmiddleware, along with providing more effi- switching is the natural evolution of net-cient Web server load balancing and caching. working technology and an enabling platformDirectory-enabled networks will radically sim- for next-generation applications. This progres-plify the management paradigm using Layer 3 sion represents the erosion of networkingswitching as a key delivery mechanism. VPNs complexity, backed by increasing performancewill become more tightly coupled with the and decreasing cost. A Layer 3 switch turnsenterprise, interfacing more closely to the out to be a well-known technology, not someLayer 3 infrastructure. VPNs will have an entirely new model. But let the buyer beware.increasingly significant role within the corpo- What looks like a true Layer 3 switch may notrate intranet, requiring more security capabili- be one at all, so it is safest to invest in a prod-ties in the Layer 3 switch. uct that was born as a true Layer 3 switch. With the advent of Layer 3 switching, theConclusion network is no longer a “Temple of Doom.”With a bit of knowledge, the Layer 3 jungle Instead, it can fulfill its promise as a key ele-doesn’t look so bad after all. In fact, Layer 3 ment of enterprise business success. Pulse Supply - Toll Free: 1-888-785-7393 - Intl: 1-951-694-1173 13