Omni product guide

download Omni product guide

of 33

Transcript of Omni product guide

  • 7/30/2019 Omni product guide

    1/33

    LARGESCALE LAN EMULATION NETWORKS

    Document Number TR 29.2346

    Cedell Alexandera

    Matt Squirea

    Rustom Masalawalab

    Seyhan Civanlarb

    aInternational Business Machines Corporation

    Research Triangle Park, N.C.

    bAT&T

    Holmdel, NJ

    January 1998

  • 7/30/2019 Omni product guide

    2/33

    ii

  • 7/30/2019 Omni product guide

    3/33

    iii

    ABSTRACT

    The ATM Forums LAN Emulation (LANE) Specification has now been available for more than two

    years, and LANE products have matured sufficiently that enterprises have deployed production

    LAN Emulation networks. However, the scaleability of LANE is still frequently debated, as is its

    applicability in WideArea Networks (WANs). This paper examines design features and supporting

    applications that can enhance the scaleability and manageability of LAN Emulation networks. The

    design features include techniques for distributing the LANE Services to provide load balancing and

    robustness, mechanisms for managing broadcast and multicast traffic (which have been a classical

    problem with large LAN networks), approaches for controlling signalling rates during powerup

    and failover situations, and extensions that exploit the distributed routing capabilities provided by

    the Next Hop Resolution Protocol (NHRP). Issues associated with WAN deployment of LAN

    Emulation are also explored, including requirements for carrierbased service environments, where

    conservation of network resources, security capabilities, and monitoring applications are especially

    important. The paper is concluded with an assessment of LANEs status/prospects as a technology

    for building largescale networks.

    ITIRC KEYWORDS

    Local Area Networks (LAN)

    Wide Area Networks (WAN)

    Asynchronous Transfer Mode (ATM)

    LAN Emulation (LANE)

    Emulated LANs (ELAN)

    ATM Forum

    Broadcast Management

    Next Hop Resolution Protocol (NHRP)

  • 7/30/2019 Omni product guide

    4/33

    iv

  • 7/30/2019 Omni product guide

    5/33

    v

    ABOUT THE AUTHORS

    Cedell Alexander is a Senior Engineer in IBMs Networking Division who is currently workingon MPOA development for IBMs Multiprotocol Switched Services (MSS) products. He was the

    chiefarchitect for Release 1.1 of the MSS Server, and led the LAN Emulation Service development

    team for Release 1.0 ofthe MSS Server. Cedell has 17+ years of industry experience, has authored

    44 technical publications along with 13 patent applications, and is a contributing member of the

    ATM Forums LAN Emulation/MPOA SubWorking Group. He received BS and MS degrees in

    Electrical Engineering from the University of Houston, and a PhD in Computer Engineering from

    Mississippi State University. Cedell may be reached by telephone at 9192542106, or via the

    Internet at [email protected].

    Matt Squire is an Advisory Engineer in IBMs Networking Division who is currently leading the

    LAN Emulation development team for IBMs Multiprotocol Switched Services (MSS) products.

    Matt developed the LAN Emulation Configuration Server for Release 1.0 of the MSS Server , and

    is a contributing member of the ATM Forums LAN Emulation/MPOA SubWorking Group. He

    received a PhD in Computer Science from North Carolina State University in 1995, and his research

    has lead to several publications. Matt may be reached by telephone at 9192548183, or via the

    Internet at [email protected].

    Rustom Masalawala is employed by AT&T in Holmdel NJ and is a contributing member of the

    ATM Forums LAN Emulation/MPOA SubWorking Group. Rustom may be reached by telephone

    at 9089497348, or via the Internet at [email protected].

    Seyhan Civanlar is employed by AT&T in Holmdel NJ and is a contributing member of the ATM

    Forums LAN Emulation/MPOA SubWorking Group. Rustom may be reached by telephone at

    9089490969, or via the Internet at [email protected].

  • 7/30/2019 Omni product guide

    6/33

    vi

  • 7/30/2019 Omni product guide

    7/33

    vii

    CONTENTS

    ABSTRACT iii

    ITIRC KEYWORDS iii. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

    ABOUT THE AUTHORS v

    LIST OF ILLUSTRATIONS ix

    ACKNOWLEDGMENTS xi

    1. INTRODUCTION 1

    2. OVERVIEW OF LAN EMULATION 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

    3. SCALEABILITYENHANCING DESIGN FEATURES 5. . . . . . . . . . . . . . . . . . . . . . .3. 1. DISTRIBUTED LANE SERVICES 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

    3. 2. BROADCAST MANAGEMENT 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3. 3. MANAGING IP MULTICAST TRAFFIC 9. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3. 4. CONTROLLING SIGNALLING RATES 11. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

    3. 5. DISTRIBUTED ROUTING 12. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

    4. WAN DEPLOYMENT ISSUES 15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4. 1. SECURITY ISSUES 17. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4. 2. MONITORING ISSUES 18. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

    5. CONCLUSION 19. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

    REFERENCES 21. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

  • 7/30/2019 Omni product guide

    8/33

    viii

  • 7/30/2019 Omni product guide

    9/33

    ix

    LIST OF ILLUSTRATIONS

    Figure 1. Physical and Logical Views of a Simple LAN Emulation Network 4. . . . . . . . . .

    Figure 2. Distributing the LANE Services with Shortcut Bridging 7. . . . . . . . . . . . . . . . . . .

    Figure 3. Managing Broadcasts at the BUS 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

    Figure 4. Managing Broadcasts at the Shortcut Bridge 9. . . . . . . . . . . . . . . . . . . . . . . . . . . .

    Figure 5. Managing IP Multicasts with Dedicated ELANs 10. . . . . . . . . . . . . . . . . . . . . . . . .

    Figure 6. OneHop Routing in a LANE Environment 13. . . . . . . . . . . . . . . . . . . . . . . . . . . .

    Figure 7. ZeroHop Routing with Route Switching 14. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

    Figure 8. WAN Configuration with Single LES/BUS Architecture 15. . . . . . . . . . . . . . . . . .

    Figure 9. WAN Configuration with Distributed LES/BUS Architecture 16. . . . . . . . . . . . . . .

    Figure 10. WAN Configuration with Distributed LES/BUS and COBased Services 16. . . . .

  • 7/30/2019 Omni product guide

    10/33

    x

  • 7/30/2019 Omni product guide

    11/33

    xi

    ACKNOWLEDGMENTS

    The IBM authors thank our manager, Rick Matlack, for his support, and acknowledge the

    contributions of the entire Multiprotocol Switched Services development team, whose long hours

    of work have made concepts described in this paper a reality.

  • 7/30/2019 Omni product guide

    12/33

    xii

  • 7/30/2019 Omni product guide

    13/33

    1

    1. INTRODUCTION

    The size ofLocal Area Networks (LANs) has traditionally been limited by two primary factors:

    physical link characteristics and broadcast traffic. A single physical link (e.g., an Ethernet or

    TokenRing LAN segment) supports a limited number of stations, offers limited geographical

    coverage, and provides limited bandwidth. One way to extend these physical limitations is to bridge

    LAN segments together. For some protocols, such as IBMs Systems Network Architecture (SNA),

    bridging has been very effective, but for other protocols, like IP, the scaleability of bridged networks

    is constrained by the volume of broadcast traffic. Consequently,routers have often been employed

    to limit broadcast domains. While routers can solve this problem, they introduce complications of

    their own, such as administrative overhead, increased latency, and potential throughput bottlenecks

    that are compounded by a price/performance ratio that tends to increase with router performance.

    Asynchronous Transfer Mode (ATM) technology, coupled with theLAN Emulation (LANE) standard

    developed by theATM Forum [1], provide the foundation for a costeffective alternative solution

    based on simple, highspeed hardware switching mechanisms.

    ATM was designed to be a universal networking technology and its benefits have been

    wellchronicled: lowcost hardware that supports scaleable networks composed of

    highbandwidth, lowlatency links with Quality of Service (QoS) guarantees. These characteristics

    make ATM appropriate for both Local and Wide Area Networks (WANs) that support a variety of

    different applications including voice, data, and multimedia traffic. Hardware switching is one of

    the fundamental components that enables ATM to offer these features, and the advantages of

    hardware switching have also been widely realized in traditional LAN environments. Together,

    ATM and LAN switching provide the building blocks for a highperformance networking

    infrastructure supporting current and future needs. The ATM Forums LAN Emulation

    Specification integrates these components in a manner enabling multivendor interoperability.

    LAN Emulation protocols allow ATM networks to provide the appearance of Ethernet andTokenRing Local Area Networks. Although LAN Emulation does not exploit all of ATMs

    capabilities, it is useful in migrating to ATM technology and lowering network management costs.

    Highspeed ATM links can be utilized while still protecting software and hardware investments.

    Software investments are protected because application interfaces are unchanged, and hardware

    investments are protected by LAN Switches that bridge traditional LANs and Emulated LANs

    (ELANs), allowing continued utilization of existing adapters/wiring. LAN Emulation also enables

  • 7/30/2019 Omni product guide

    14/33

    2

    an incremental migration strategy where ATM adapters are gradually installed in stations requiring

    additional bandwidth (e.g., servers and engineering/multimedia workstations).

    The network management benefits of ELANs come from increased configuration flexibility.

    Membership in an Emulated LAN is not based on physical location; instead, logicallyrelated

    stations are grouped to form an ELAN. As long as ELAN memberships are retained, no

    reconfiguration is needed when stations move to new physical locations. Similarly, no wiring

    modifications are needed to move stations from one ELAN to another.

    When augmented with the design features and supporting applications described in this paper, the

    advantages of LAN Emulation can be extended to largescale enterprise networks. A brief

    overview of LAN Emulation basics is provided in Section 2. as a preface to the discussion of

    enhancements which improve the scaleability of LANE in Section 3. Topics covered in Section3. include techniques for distributing the LANE Services, mechanisms for managing broadcast and

    multicast traffic, approaches for controlling signalling rates, and extensions that exploit the

    capabilities provided by the Next Hop Resolution Protocol (NHRP). Issues associated with WAN

    deployment of LAN Emulation are examined in Section 4. with a focus on requirements for

    carrierbased service environments, where conservation of network resources, security capabilities,

    and monitoring applications are particularly important. The paper is then concluded in Section

    5. with an assessment of LANEs status/prospects as a technology for building largescale networks.

  • 7/30/2019 Omni product guide

    15/33

    3

    2. OVERVIEW OF LAN EMULATION

    The components that implement an ELAN are LAN Emulation Clients (LECs), the LAN Emulation

    Configuration Server (LECS), the LAN Emulation Server (LES), and the Broadcast and Unknown

    Server (BUS). The LES, BUS, and LECS are collectively referred to as the LE Service components.

    Each ELAN has a dedicated LES and BUS (but a single hardware platform can provide LE Services

    for multiple ELANs). LECs reside in end systems (e.g., ATMattached hosts or LAN switches

    representing hosts residing on traditional LAN segments) and provide a Media Access Control

    (MAC)level service to higherlayer software. To emulate a LAN, LECs request services from the

    LECS, LES, and BUS.

    The LECS reduces the network management burden by serving as a centralized repository for

    configuration data, minimizing configuration at LECs. The LECS helps create aplugandplay

    environment by providing initial configuration data to LECs, with the most crucial piece of data

    being the ATM address of the LES. LECs connect to the LECS using welldefined procedures that

    include obtaining the address of the LECS from the ATM switch or use of a WellKnown LECS

    ATM Address.

    After receiving the ATM address of the LES, the LEC begins to join the ELAN by setting up a

    Control Direct Virtual Channel Connection (VCC) to the LES and sending anLE_JOIN_REQUEST. The LES typically responds by adding the LEC to a pointtomultipoint

    Control Distribute VCC and returning a successful LE_JOIN_RESPONSE. The LEC then issues

    an LE_ARP_REQUEST for the all 1s broadcast MAC address to obtain the ATM address of the

    BUS. When the LE_ARP_RESPONSE is returned, the LEC establishes a Multicast Send VCC to

    the BUS, and the BUS responds by adding the LEC to a pointtomultipoint Multicast Forward

    VCC, at which time, the LEC is considered to be a member of the ELAN.

    LECs can register LAN Destinations with the LES to ensure uniqueness and allow the LES to answer

    LE_ARP_REQUESTs. Registrations include the LAN Destination and the ATM address that the

    LEC associates with the LAN Destination. A LAN Destination may be either a MAC Address, or

    a Route Descriptor if the LEC is integrated with a sourceroute bridge.

  • 7/30/2019 Omni product guide

    16/33

    4

    When a LEC needs to send a unicast frame to a LAN Destination for which it does not have a direct

    connection, the LEC sends the frame to the BUS on the Multicast Send VCC. The LEC also sends

    all multicast/broadcast frames to the BUS. The BUS is responsible for distributing these frames to

    the appropriate destination(s). One simple way for the BUS to deliver these frames is to forwardthem to all LECs on a Multicast Forward VCC. More efficient approaches are described in Section

    3. 2.

    The LEC monitors the amount of unicast data sent to each LAN Destination through the BUS, and

    when the traffic rate is sufficient to warrant direct communication with a particular LAN

    Destination, the LEC initiates address resolution by transmitting an LE_ARP_REQUEST to the

    LES. If the LAN Destination is registered, the LES responds with the associated ATM address.

    Otherwise, the LES forwards the request on a Control Distribute VCC, and the LEC representing

    the LAN Destination responds with an appropriate ATM address. Upon receiving the

    LE_ARP_RESPONSE, the initiating LEC uses the ATM address to establish a Data Direct VCC,

    and subsequent traffic to the LAN Destination is transmitted directly over this VCC.

    Figure 1 below completes the overview by depicting physical and logical views of a simple LAN

    Emulation network.

    ATMSWITCH

    LE Service

    LECS LESa,b BUSa,b

    LEC

    LANDeviceDriver

    Applications

    ATMHost1 ELANa

    LEC LEC

    Bridging Function

    Legacy Host1 ELANa Legacy Host2 ELANb

    LANSWITCH

    ATM Host2 ELANb

    PHYSICAL NETWORK LOGICAL NETWORK

    ELANa

    ATMHost1

    Legacy Host1

    ELANb

    ATM Host2

    Legacy Host2

    Figure 1. Physical and Logical Views of a Simple LAN Emulation Network

  • 7/30/2019 Omni product guide

    17/33

    5

    3. SCALEABILITYENHANCING DESIGN FEATURES

    3. 1. DISTRIBUTED LANE SERVICES

    The centralized aspects of the LE Service can be both advantageous and problematic. For example,

    the LES/BUS represents a potential singlepointoffailure for an ELAN. Although the LANE

    v1.0 Specification does not address LES/BUS redundancy, a number of vendors have recognized

    that this singlepointoffailure is unacceptable in many production networks, and have added

    extensions supporting failover to a backup LES/BUS [2]. Robustnessenhancing extensions of this

    type have greatly aided customer acceptance of LANE products. Naturally, the next logical step is

    to distribute the LE Service components in order to realize scaleability benefits along with the

    increased reliability.

    With a distributed LE Service, multiple LES/BUSs are available to service the clients of a single

    ELAN. Thus, reliability is enhanced, since the clients of a failed LES/BUS may be reassigned to

    any of the remaining LES/BUSs. Distributing the LE Services also addresses a number of

    resourcerelated scaleability issues that can hinder deployment of largescale LANE networks. A

    single LES/BUS instance may be limited in terms of processing power, forwarding throughput,

    memory availability, or VCC capacity. However, with multiple LES/BUSs serving an ELAN, the

    cumulative resources can be scaled up to meet the needs of the installation. Additionally,distributing the LE Services can decrease client join times during highactivity periods, such as

    poweron sequences, by partitioning the signalling load across the ATM network.

    In response to user requirements for these features, the ATM Forum is currently defining a standard

    for distributing the LE Service. This standard is called the LANE NetworkNetwork Interface

    (LNNI) Specification [3]. LNNI defines a protocol between LE Service components that enables

    multiple LES/BUSs to cooperatively serve a single ELAN. LNNIenhanced servers will establish

    VCCs explicitly for servertoserver communication, and exchange control and data frames over

    these VCCs so that clients joined to different LESs may communicate directly (as if they were joined

    to the same LES).

  • 7/30/2019 Omni product guide

    18/33

    6

    The LNNI Specification is currently in draft form; consequently, LNNI implementations may not

    be available for quite some time. Additionally, existing LANE v1.0 servers will require software

    upgrades to be able to participate in the LNNI protocol. To satisfy current requirements for

    distributed LANE Services while protecting investments in existing technology, a new type ofinterconnection device may be deployed. This device implements Shortcut Bridging [4], an

    enhanced version of transparent bridging.

    When ELAN segments are bridged together using conventional techniques, independent LES/BUS

    instances service each ELAN segment, and stations on different ELANs communicate by

    establishing Data Direct VCCs to the bridge, which forwards interELAN traffic. Thus, conventional

    ELAN bridging techniques can distribute the LE Service in a robust manner (since there are multiple

    LES/BUS instances serving the collective set of clients and standard spanning tree protocols protect

    against failure of the bridge itself). However, all interELAN traffic must pass through the bridge,

    creating a potential bottleneck in terms of forwarding power and VCC resources. Shortcut bridging

    solves this problem by enabling Data Direct VCCs to be established between devices residing on

    different ELAN segments as illustrated in Figure 2 below.

    To better understand shortcut bridging operation, suppose that Stationx in Figure 2 wishes to

    communicate with Stationy, and that Stationx has issued an LE_ARP_REQUEST for the MAC

    address ofStationy. Bridging LECx receives the LE_ARP_REQUEST, and the shortcut bridge either

    answers the request (if it has a cache entry for the destination), or forwards the request to the LES

    for ELANy through Bridging LECy. Assume LESy answers the LE_ARP_REQUEST. The

    LE_ARP_RESPONSE is received byBridging LECy, and forwarded toLESx byBridging LECx.

    LESx forwards the LE_ARP_RESPONSE on to Stationx, and Stationx then sets up a Data Direct

    VCC to Stationy.

    The fundamental difference between the operation of a shortcut bridge and that of a conventional

    ELAN bridge is that the shortcut bridge returns the ATM address of the destination station, while

    a conventional ELAN bridge would have returned the ATM address ofBridging LECx in responsetoStationxs LE_ARP_REQUEST. This is however a significant difference, since it preserves the

    essence of belonging to the same ELAN by permitting clients to communicate directly over the

    switched ATM network, and thereby enables the collection of bridged ELANs to function as a single

    SuperELAN. In summary, shortcut bridging is a robust control mechanism for distributing the LE

    Services in a standardsbased manner that builds upon proven characteristics of traditional bridging

    techniques while eliminating the performance bottlenecks.

  • 7/30/2019 Omni product guide

    19/33

    7

    Bridging

    LECx

    Shortcut Bridge

    ELANxLES/BUSx

    Bridging

    LECy

    ELANy LES/BUSy

    Shortcut Data Direct VCC

    StationxStationy

    Figure 2. Distributing the LANE Services with Shortcut Bridging

    3. 2. BROADCAST MANAGEMENT

    The propagation of broadcast traffic has been a deterrent to the deployment of largescale LANs in

    many environments. Broadcast traffic brings the twin evils of consuming network bandwidth on

    every LAN segment and creating perturbations at all end stations. Fortunately, these difficulties can

    be managed in LAN Emulation networks by leveraging the centralized aspects of the LE Service,

    which are an advantage when it comes to controlling broadcasts.

    Since the BUS forwards all broadcast traffic, it is an ideal platform for intelligent broadcast

    management functions. Broadcast management techniques are effective because most broadcast

    frames are not actually intended for all stations. IP Address Resolution Protocol (ARP) requests are

    a prime example. IP ARPs are intended for a particular station with an unknown MAC address.

    For the ARP protocol to work, it is only necessary that the station with the specified IP address

    receive the ARP request. Examples from other protocols include NetBIOS Name Queries, which

    are conceptually similar to IP ARPs, and IPX RIP/SAP advertisements, which are only intended for

    IPX routers and servers.

    Figure 3 illustrates two broadcast management techniques, unicast conversion and directed

    transmission, that can enhance the scaleability of LANE networks by reducing both network trafficand end station disturbances. In this example, the BUS has been augmented with aBroadCast

    Manager (BCM) [2] component that transforms selected broadcasts into unicast frames. The

    transformed frames are then transmitted directly to the destination LANE Client, and not to all

    LANE clients. BCM operates by dynamically learning and caching the {protocol address, MAC

    address, VCC} associations present in the network. To see how this works, lets examine the steps

    depicted in Figure 3.

  • 7/30/2019 Omni product guide

    20/33

    8

    1) Assume the BCM IP cache already contains a mapping for the IP address ofStationy, and

    that Stationx transmits an IP ARP request forIPy to the BUS.

    2) The BUS recognizes the broadcast MAC address and passes the frame to BCM. BCM

    examines the header and identifies the frame as an IP ARP request. BCM then obtains thecached entry for IPy and inserts a new cache entry for IPx. The cache entry maps the IP

    address to the corresponding MAC address and identifies the Multicast Send connection to

    the LEC representing the address.

    3) Using the cached information, BCM converts the IP ARP request to a unicast frame by

    inserting the proper destination MAC address, and transmits the frame directly to the

    appropriate LEC on a Multicast Send VCC.

    4) The unicast conversion extends BCM benefits to traditional LANs when the destination

    station is located behind a bridge function.

    Stationx Stationy

    BUS

    BCM

    IP Cache

    IPy, MACy, VCCyIPx, MACx, VCCx...

    LAN

    SWITCH

    Unicast IP ARP Request(Dst MAC = MACy,Src MAC = MACx,Dst IP = IPy,Src IP = IPx)

    Broadcast IP ARP Request(Dst MAC = all 1s,Src MAC = MACx ,Dst IP = IPy,Src IP = IPx)

    2

    13

    4

    Multicast Send VCCx Multicast Send VCCy

    Figure 3. Managing Broadcasts at the BUS

    BCM techniques are not limited to the BUS, and may be effective at any network junction that

    processes a large amount of broadcast frames. For example, the techniques can be applied at a

    shortcut bridge to create a broadcast management hierarchy, with BCM managing broadcasts within

    an ELAN segment and a Bridging BroadCast Manager (BBCM) [4] component managing

    interELAN broadcasts. Dynamic Protocol Filtering (DPF) [4] is another, complementary broadcastcontrol mechanism that can be utilized at shortcut bridges. DPF dynamically partitions the network

    into Protocolspecific Virtual LANs (PVLANs) by monitoring the traffic received from each ELANto learn the protocols and subnets being used on that segment. Each PVLAN constitutes a broadcast

    domain; thus, the scope of unmanaged broadcast/multicast packets for a particular protocol/subnet

    is reduced to those segments that are actually utilizing the protocol/subnet. Figure 4 shows an

    example configuration with four PVLANs: IP Subnet 1,IP Subnet 2 ,IPX Network 1 , andNetBIOS.

    In this example, IP ARP traffic for destinations inIP Subnet 1 is restricted toELANa andELANc.

    Likewise, IPX and NetBIOS traffic is also restricted to the indicated subset of ELANs.

  • 7/30/2019 Omni product guide

    21/33

    9

    BridgingLECa

    PVLAN Shortcut Bridge

    IP Subnet 1IP Subnet 2

    BridgingLECc

    IPX Net 1NetBIOS

    IP Subnet 1

    BridgingLECb

    BridgingLECd

    IP Subnet 2

    BridgingLECf

    IPX Net 1 NetBIOS

    BridgingLECe

    BBCM

    ELANa ELANb ELANc ELANd ELANe ELANf

    Figure 4. Managing Broadcasts at the Shortcut Bridge

    3. 3. MANAGING IP MULTICAST TRAFFIC

    Driven by Internetrelated developments, an increasing number of useful applications are emerging

    that rely on IP multicast protocols to support multiple senders and receivers for purposes such as

    collaborative computing and multimedia distribution. As these applications grow, so do the

    demands on the network to provide techniques for efficiently delivering multicast traffic. The

    traditional approach has been for hosts to advertise their interest in participating in a particular

    multicast group using theInternet Group Management Protocol (IGMP) [5], and for routers to then

    form a distribution tree using a multicast routing protocol, such as theDistance Vector Multicast

    Routing Protocol (DVMRP) [6] orMulticast OSPF (MOSPF) [7].

    IP multicast packets are transmitted with layer2 multicast addresses on LAN media, and if no

    management techniques are employed within an ELAN, the multicast traffic will be forwarded to

    all LECs by the BUS, with resulting problems akin to those associated with unmanaged broadcasts.

    One approach to solving this problem is to add IP multicast management intelligence to the BUS.

    The BUS could snoop IGMP messages in order to learn which LECs are supporting which IP

    multicast groups, and then only forward multicast packets to the appropriate subset of LECs. Similarintelligence could also be added to the shortcut bridge to createIP Multicast PVLANs. A positive

    attribute of this approach is that the support for managing IP multicast traffic is isolated within the

    components providing the distributed LE Services, and no changes are required at the LECs.

    However, the approach does not make optimal utilization of ATM network capabilities, as multicast

    packets are replicated within the LE Service components, instead of simply being transmitted on a

    pointtomultipoint VCC, which is ideally suited for delivering multicast traffic.

  • 7/30/2019 Omni product guide

    22/33

    10

    An approach that better utilizes the capabilities of the ATM network is illustrated in Figure 5 below.

    Here, ELANs have been dedicated to particular IP multicast address ranges, and the LAN switches

    have been enhanced to direct multicast traffic to the appropriate ELAN (i.e., only multicast traffic

    for the associated group(s) flows on the dedicated ELANs). By also incorporating IGMP snoopingtechniques, the LAN switches can confine multicast traffic received from an ELAN to the LAN

    segments containing hosts that are currently participating in the multicast group, which makes for

    an efficient solution. Note that in this configuration, no special functions are required at the BUS.

    The BUS simply forwards received traffic on a pointtomultipoint Multicast Forward VCC to all

    the LECs in the ELAN, which happen to all be supporting members of IP multicast groups; thus,

    no network bandwidth is wasted and no stations are unnecessarily disturbed. This simplicity of

    function also makes it much easier to build hardware enabling linerate forwarding at the BUS.

    Furthermore, the Multicast Forward VCCs established by these BUSs can be setup with QoS

    parameters matched to the traffic requirements of the corresponding multicast groups, and the LECs

    and LES/BUSs required for new Multicast ELANs can be created dynamically on an asneeded

    basis.

    BUSx

    LEC LEC

    LANSWITCH1

    Groupz230.0.5.3

    Groupx224.0.1.7

    . . . LEC LEC

    LANSWITCH2

    Groupz230.0.5.3

    Groupx224.0.1.7

    . . . LEC LEC

    LANSWITCHN

    Groupz230.0.5.3

    Groupx224.0.1.7

    . . .. . .

    BUSz. . .

    Multicast Forward VCC (CBR)1.5 Mbps

    Multicast Forward VCC (rtVBR)PCR = 5 Mbps

    SCR = 500 Kbps

    Multicast Address Ranges

    ELANx: 224.0.1.5 229.0.5.10. . .

    ELANz: 230.0.5.1 2234.0.5.10

    Figure 5. Managing IP Multicasts with Dedicated ELANs

    Version 2.0 of the LANE UserNetwork Interface (LUNI) Specification [8] helps with multicasttraffic management by specifying procedures that standardize the fundamental approach of

    dedicating LE Service forwarding entities to multicast groups. LANE v2.0 LECs issue

    LE_ARP_REQUESTs for multicast MAC addresses that they wish to transmit to, and are capable

    ofaccepting multiple Multicast Forward VCCs. The LE_ARP_REQUEST provides the LES with

    an opportunity to return the ATM address of a BUS that is dedicated to the multicast group, while

    the capability of establishing multiple Multicast Forward VCCs provides the BUS considerable

  • 7/30/2019 Omni product guide

    23/33

    11

    flexibility in limiting the scope of multicast transmissions. In addition to these basic changes, LUNI

    v2.0 also introduces an optional Selective Multicastfunction. LECs that support Selective Multicast

    register the multicast MAC addresses for which they wish to receive traffic with the LES. This

    registration allows the BUS to limit the multicast traffic sent to the LEC without performing anyIGMP snooping.

    When compared with traditional routerbased approaches, LANEbased multicasting systems offer

    several key advantages:

    1) With LANEbased solutions, multicast traffic is logically separated from unicast traffic,

    and all multicast packet forwarding decisions can be performed at layer2, even when the

    destination is in a different subnet. This improves throughput and lowers latency by

    replacing routing with switching on the multicast data path.

    2) The procedures employed by multicast routing protocols to determine the properdistribution trees generate significant overhead. This is almost completely absent in a

    LANEbased solution, thereby making better use of the available bandwidth.

    3) The pointtomultipoint circuits that are built from the BUS to the LECs can be setup with

    the desired Quality of Service characteristics, which is an important enabler for efficient

    and easily managed QoSbased applications.

    3. 4. CONTROLLING SIGNALLING RATES

    Events such as poweron sequences and server failover can create situations where a large number

    ofclients are simultaneously attempting to connect to the same LE Service component, which can

    lead to unusually high signalling rates at the ATM switch providing access to the target LE Service

    component. This can result in overflowed queues and delayed, if not denied, network service

    because of the loss of signalling messages. Experience with LANE v1.0 has shown that

    implementation of a random retry time interval at the LEC is quite effective in controlling signalling

    rates and solving this problem. Unfortunately, use of a random retry time interval is not mandated

    in the LANE v1.0 Specification (although it has been in the LUNI Version 2.0 Specification). Thus,

    it is useful for the LES/BUS to also implement a signalling congestion control algorithm.

    The LES/BUS can help control the signalling load by pacing the rate at which clients are added to

    its pointtomultipoint connections when the ATM network is congested. If congestion is detected

    (via cause codes in signalling messages), the LES/BUS should continue to accept pointtopoint

    VCCs (i.e., Control Direct and Multicast Send VCCs), since rejecting these VCCs would only add

    to the congestion, but delay adding the clients to the pointtomultipoint VCCs (i.e., Control

    Distribute and Multicast Forward) for a random interval using an exponential backoff algorithm.

  • 7/30/2019 Omni product guide

    24/33

    12

    This temporal distribution of the signalling load lowers the probability of lost messages and thereby

    expedites service restoration. Such an algorithm is required in largescale networks with LECs that

    do not implement a random retry interval, and is advantageous in any case.

    3. 5. DISTRIBUTED ROUTING

    Design features supporting large ELANs have been described in the preceding sections. As ELANs

    become larger, networks become flatter, which offers the dual benefits of simplicity and more

    efficient utilization of highspeed switching infrastructures. When the number of subnets is

    reduced, configuration is simplified, which naturally eases network management. Furthermore,

    larger subnets tend to increase intrasubnet traffic within the enterprise, resulting in more direct

    switched connections, with a corresponding reduction in routing bandwidth requirements for

    intersubnet traffic. Nonetheless, there are still situations where routing functions are needed. For

    example, security concerns or existing addressing conventions may dictate a more highly subnetted

    design. TheNext Hop Resolution Protocol (NHRP) [9] was designed for these situations.

    NHRP is an Internet Engineering Task Force (IETF) protocol that can improve network

    performance by eliminating router hops on NonBroadcast MultiAccess (NBMA) subnetworks

    like ATM. When internetworking protocols, such as IP, are run over ATM subnetworks, the routed

    path may include multiple hops across the same ATM subnetwork. NHRP allows better utilization

    of the underlying switched infrastructure by enabling the establishment of shortcut routes across

    ATM subnetworks.

    NHRP is a clientserver protocol. NHRP Clients (NHCs) issue requests to NHRP Servers (NHSs),

    and NHSs either respond to the requests or forward the requests along the routed path. To initiate

    establishment of a shortcut to a particular destination, an NHC transmits an NHRP Resolution

    Request along the routed path. The NHS that serves the destination (e.g., when the destination

    resides on a subnet local to the NHS) responds with an ATM address to be used in setting up a shortcut

    VCC. In the bestcase, NHRP enableszerohop routing, where all router hops are removed from

    the steadystate data path. Additionally, since the resolution requests follow the routed path,routerbased security capabilities are preserved in NHRP environments (as protocollayer access

    controls may be applied to the resolution requests).

    Although the NHRP Specification does not define support for LANE encapsulation, it does define

    a generic, interoperable extension mechanism that can be readily utilized to extend the benefits of

    NHRP to the large installed base of LANE equipment. NHRP support for LANE encapsulation is

  • 7/30/2019 Omni product guide

    25/33

    13

    particularly useful in onehop routing, where the ingress router establishes shortcuts on behalf of

    hosts without NHCs as depicted in Figure 6 (i.e., whereRouter1 setups up an intersubnet shortcut

    toLAN Switch2 on behalf ofIP Hostx for traffic destined toIP Hosty). Onehop routing is attractivebecause no end system changes are required.

    ATM

    LANSwitch1

    Ethernet ELAN

    IP Subnet9.1.3

    IP Subnet9.1.2

    TokenRing ELANIP Subnet

    9.1.1

    NHRP/Router1

    LEC1

    LAN Switch2

    LEC2

    IP Hosty

    9.1.3.1IP Hostx

    9.1.1.1

    IP Hostw

    NHRP/Router2

    LEC

    Intersubnet Shortcut LANE Data Direct VCCs

    Figure 6. OneHop Routing in a LANE Environment

    NHRP support for LANE encapsulation also provides the basis for extending the benefits of

    zerohop routing to stations residing on traditional LAN segments throughRoute Switching. Route

    Switching is the ultimate form of edge routing, with layer3 forwarding decisions distributed across

    LAN stations. To achieve this, NHC function is included in the LAN stations. Route Switching

    Clients dynamically learn the MAC addresses of NHSs and monitor the traffic flows destined for

    these MAC addresses. The rate of traffic to a particular layer3 destination can then trigger ashortcut request. A shortcut is requested by sending an NHRP Resolution Request to the NHS.

    These requests use the MAC address family as opposed to the ATM address family. The MAC

    address associated with the destination protocol address is provided in the NHRP Resolution Reply.

    Upon receiving the reply, the Route Switching Client substitutes the destination MAC address for

    the routers MAC address and frames are then switched endtoend bypassing all intermediate

    routers as shown in Figure 7.

  • 7/30/2019 Omni product guide

    26/33

    14

    Intersubnet Shortcut Data Direct VCC

    ATM

    LAN SwitchLAN Switch

    ELAN IP Subnet

    9.1.3.0

    ELAN IP Subnet

    9.1.2.0ELAN IP Subnet

    9.1.1.0

    LAN IP Subnet

    9.1.1.0

    IP HostA IP HostB

    ProtocolStack

    ProtocolStack

    Like Media(both Ethernet or both TokenRing)

    NHRP ClientFunction

    NHRP ClientFunction

    NHS/Router NHS/Router

    LAN IP Subnet

    9.1.1.0

    Figure 7. ZeroHop Routing with Route Switching

  • 7/30/2019 Omni product guide

    27/33

    15

    4. WAN DEPLOYMENT ISSUES

    ATM provides the capabilities to easily extend an ATM LAN across a Wide Area Network given

    that there are essentially no distance limitations inherent in ATM as a data link layer protocol. This

    property of ATM permits Local Area Networks to be extended across a WAN at high speeds, such

    as DS3 and OC3, allowing workgroups to be formed that span many different sites while emulating

    a single LAN using ATM switching. At these speeds, there will be no discernible difference between

    communicating with local and remote stations on the LAN.

    Although LANE v1.0 provides all the capabilities needed to stretch an ELAN across the WAN,

    single LES/BUS architectures are not the most efficient solution for WAN environments, where

    network resources are more precious. For example, consider the network shown in Figure 8 with

    four sites spread across the United States. With the LE Services located in Seattle, four VCCs across

    the WAN are required to connect each of the LECs in the other sites to the LES/BUS, and address

    resolution requests generated by these LECs will need to traverse the WAN, even when the target

    station is located at the same site as the originator of the request.

    Seattle

    LES/BUS

    LECs

    DallasLECs

    Raleigh

    LECs

    BostonLECs

    Figure 8. WAN Configuration with Single LES/BUS Architecture

    The distributed LE Service architecture depicted in Figure 9 eliminates the inefficiencies described

    above by deploying a LES/BUS at each site. In this solution, the WAN is still used as a simple

    transport network providing Switched Virtual Circuit (SVC) connectivity between LANE

    components across the multiple sites. However, additional robustness and efficiencies can be

  • 7/30/2019 Omni product guide

    28/33

    16

    obtained by adding CentralOffice (CO) based LANE Services to the ELAN as illustrated in

    Figure 10. The CObased servers are owned and maintained by a telecommunications

    administrator, which is economical, since the same equipment can be reused for many different

    customers. Thus, with secure partitioning between the ELAN domains, CObased services cansupport ELANbased Virtual Private Networks (VPNs).

    Seattle

    LES/BUS

    LECs

    Dallas

    Raleigh

    Boston

    LES/BUS

    LECs

    LES/BUS

    LES/BUSLECs

    LECs

    Figure 9. WAN Configuration with Distributed LES/BUS Architecture

    Seattle

    LES/BUS

    LECs

    Dallas

    Raleigh

    Boston

    LES/BUS

    LECs

    LES/BUS

    Central OfficeLANE Services

    LES/BUS

    LES/BUS Interconnection,Broadcast/Multicast Management

    LECs

    LECs

    Figure 10. WAN Configuration with Distributed LES/BUS and COBased Services

  • 7/30/2019 Omni product guide

    29/33

    17

    One obvious advantage of the CO model over mesh interconnection is that, for an ELAN with n sites,

    the number of LE ServicetoLE Service connections is reduced from O(n2) to O(n). CO servers

    can achieve this reduction now by interconnecting the LES/BUSs with a shortcut bridge, and offer

    a migration path to LNNI later when the standard is completed. CO servers can also provideLES/BUS facilities for small sites where it would be inefficient to support a local LES/BUS.

    Additionally, CO servers are a convenient platform for deploying intelligent broadcast and multicast

    management functions that preserve valuable WAN bandwidth. Similarly, CO services can be easily

    extended to include an NHRPbased routing offering. Along with a distributed architecture

    characterized by high scaleability and carriergrade reliability, security and monitoring capabilities

    are also key attributes of central office servers.

    4. 1. SECURITY ISSUES

    To support Virtual Private Networks, CObased LANE Servers must be able to restrict data access

    to Closed User Groups. This type of security can be realized at two levels. The first of these is the

    physical layer, where ATM switches are cognizant of the permitted sourcedestination pairs, and

    connections violating the rules are not allowed to complete. While physicallayer security is

    effective, it can also be restrictive and cumbersome to administer on a largescale basis, which

    brings us to the other option, applicationlayer security.

    Applicationlayer security is obviously dependent upon the particular application. So, well use a

    couple of examples to illustrate the concepts. The first example is LES/BUS security, where the task

    is preventing unauthorized users from joining the ELAN. One basic applicationlayer security

    approach is denial of information. This approach can be applied to LES/BUS security by not

    providing the ATM address of the LES/BUS to ineligible users. Denial of information can make

    it more difficult to breach security, but is clearly insufficient in the case of CObased servers. A

    second applicationlayer approach is jointime security. Here, the LES/BUS validates clients at the

    time they join the ELAN. One way that the LES/BUS can validate clients is to query the LECSs

    ELAN assignment policy database. Use of the ATM address assignment policy, in conjunction with

    address screening at the ATM switches, is an effective combination.

    The second example involves shortcut bridging, where the task is to ensure that traffic remains

    within the appropriate ELANs. This can be accomplished in a straightforward manner by simply

    assigning each bridging LEC to a particular domain and only forwarding traffic over ports that are

    in the same domain. Filtering may also be performed at the applicationlayer, as well as the

    physicallayer. For example, many bridges include an extensive set of filtering capabilities that are

  • 7/30/2019 Omni product guide

    30/33

    18

    applied on a framebyframe basis. These same capabilities can be applied to address resolution

    requests received by shortcut bridges to prevent shortcut establishment.

    4. 2. MONITORING ISSUES

    Effective management applications are crucial to the success of largescale LAN Emulation

    networks. Simple Network Management Protocol (SNMP) [10] agents are needed at each ELAN

    component to track configuration changes, manage faults, and monitor performance levels. As

    many network elements, such as LECs and LES/BUSs, may be created or removed dynamically, a

    systemwide view is required to monitor the network configuration. Each agent generates traps

    when local configuration events occur. These traps are delivered to network management stations

    where they are collected by configuration management applications that update graphical topology

    maps depicting the current network configuration.

    In addition to the configuration traps, traps are also needed for fault management and performance

    monitoring purposes. For example, LANE failures may result in SNMP traps being delivered to

    management stations permitting centralized management of the distributed system. Management

    stations would typically respond to such a failure by changing the color of the affected network

    element in the topology display, logging a message in a history file, and possibly paging the network

    operator. Thus, the graphical displays create a visual capture of the current and past network, which

    is a significant aid to timely analysis and response in the network control center.

    Configurable triggers that generate alarms when the values of specified MIB elements cross upper

    or lower thresholds are another important feature of an effective monitoring system. These triggers

    may be based on elements in the standard MIBs developed by the ATM Forum, or on new elements

    that have been added to Enterprise MIBs to enhance network manageability. While configuration

    and fault management issues tend to be relatively straightforward, performance management is a

    more challenging task. Although its easy to collect and display performancerelated statistics, like

    the number of frames forwarded, its not easy to relate these statistics to meaningful management

    issues such as identification of users that are abusing the network or determining when additional

    capacity is needed. This is an area where additional work is needed as new tools that do a better job

    of relating performance data to capacity planning would be of substantial value.

    CObased servers also place some unique demands on the management system. One example is the

    capability for users to control the configuration of their VPNs, preferably via the web, without being

    able to impact the configuration of other users VPNs. Other examples include the ability to monitor

    on a VPN basis, and acquisition of detailed statistical information permitting accurate billing.

  • 7/30/2019 Omni product guide

    31/33

    19

    5. CONCLUSION

    It is becoming increasingly clear that ATM hardware is the technology of choice for the core of

    telecommunication carrier and Internet Service Provider (ISP) networks. The protocols that will

    run on this hardware and the role of ATM in local area networks are more controversial topics. As

    one begins to look at the issues associated with collaborative computing and multicasting, it becomes

    apparent that innovative architectures are going to be required to solve these complex problems.

    Nonetheless, while these issues are being debated, LANE implementations are maturing and a

    growing number of LANE networks are being deployed to carry multiprotocol traffic in campus

    environments. Although there are scaleability and reliability issues associated with strictimplementations of the LANE v1.0 Specification, availability of products that implement the design

    features described in this paper, coupled with economical WAN services, will enable the design of

    networks that do not suffer from these drawbacks and are capable of growth across the enterprise.

    With this momentum, LAN Emulation appears to be wellpositioned as a technology for building

    largescale networks that support both existing applications and the new class of emerging

    QoSbased applications.

  • 7/30/2019 Omni product guide

    32/33

    20

  • 7/30/2019 Omni product guide

    33/33

    REFERENCES

    [1] ATM Forum, LAN Emulation Over ATM: Version 1 Specification,

    AFLANELNNI0021.000, Jan. 1995.

    [2] C. Alexander, etal., Quick Guide to the IBM Multiprotocol Switched Services (MSS) Server

    Release 1.0, Tech. Rep. TR 29.2170, IBM, Aug. 1996 (available from

    http://www.raleigh.ibm.com/tr2/tr2over.html).

    [3] ATM Forum, LAN Emulation Over ATM Version 2 LNNI Specification Draft 8,

    BTDLANELNNI02.08, Feb. 1997.

    [4] C. Alexander, Quick Guide to the IBM Multiprotocol Switched Services (MSS) Server Release

    1.1, Tech. Rep. TR 29.2260, IBM, May 1997 (available from

    http://www.raleigh.ibm.com/tr2/tr2over.html).

    [5] S. Deering, Host Extensions for IP Multicasting, RFC 112, Aug. 1989.

    [6] D. Waitzman, C. Partridge, and S. Deering, Distance Vector Multicast Routing Protocol,

    RFC 1075, Nov. 1988.

    [7] J. Moy, Multicast Extensions to OSPF, RFC 1584, Mar. 1994.

    [8] ATM Forum, LAN Emulation Over ATM Version 2 LUNI Specification,

    AFLANE0084.000, July 1997.

    [9] J. Luciani, D. Katz, D. Piscitello, and B. Cole, NBMA Next Hop Resolution Protocol

    (NHRP), InternetDraft, .

    [10] J. Case, M. Fedor, M. Schoffstall, and J. Davin, A Simple Network Management Protocol

    (SNMP), RFC 1157, May 1990.