Chapter Six. ATM Internetworking Design

Asynchronous Transfer Mode (ATM) is the first networking architecture developed specifically for supporting multiple services. ATM networks are capable of supporting audio (voice), video and data simultaneously. ATM is currently architected to support up to 2.5 Gbps bandwidth. Data networks immediately get a performance enhancement when moving to ATM due to the increased bandwidth over a WAN. Voice networks realize a cost savings due in part to sharing the same network with data and through voice compression, silence compression, repetitive pattern suppression, and dynamic bandwidth allocation. The ATM fixed-size 53-byte cell enables ATM to support the isochronicitiy of a time-division multiplexed (TDM) private network with the efficiencies of public switched data networks (PDSN).

Most network designers are first challenged by the integration of ATM with the data network. Data network integration requires legacy network protocols to traverse a cell-based switched network. ATM can accomplish this in several ways. The first of these is LAN emulation.

  1. LAN emulation (LANE)

    ATM employs a standards based specification for enabling the installed base of legacy LANs and the legacy network protocols used on these LANs to communicate over an ATM network. This standard is known as LAN emulation (LANE). LANE uses the Media Access Control (MAC) sublayer of the OSI data link control Layer 2. Using MAC encapsulation techniques enables ATM to address the majority of Layer 2 and Layer 3 networking protocols. ATM LANE logically extends the appearance of a LAN thereby providing legacy protocols with equivalent performance characteristics as are found in traditional LAN environments. Figure 6.1 illustrates a typical ATM topology with LANE support.

    LANE can use ATM emulated LANs (ELANs).. Using ELANs, a LAN in one location is logically connected to a LAN in another location. This allows a network designer to extend a LAN over an ATM WAN avoiding the need for routing packets between the two locations. LANE services can be employed by ATM attached serves or workstations, edge devices such as switches, and routers when routing between ELANs is required. ATM LANE uses four components to establish end-to-end connectivity for legacy protocols and devices. These are LAN Emulation Client, LAN emulation configuration server (LECS), LAN emulation server (LES), and Broadcast and Unknown Server (BUS).

    1. LAN Emulation Client (LEC)

    Any end system that connects using ATM require a LAN emulation Client (LEC). The LEC performs the emulation necessary in support of the legacy LAN. The functions of the LEC are:

    • Data forwarding
    • Address resolution
    • Registering MAC addresses with the LANE server
    • Communication with other LECs using ATM virtual channel connections (VCCs).

    End systems that support the LEC functions are:

    • ATM-attached workstations
    • ATM-attached servers
    • ATM LAN switches (Cisco Catalyst family)
    • ATM attached routers (Cisco 12000, 7500, 7000, 4700, 4500 and 4000 series)
        1. LAN Emulation Configuration Server (LECS)

    The ELAN database is maintained by the LAN emulation configuration server (LECS). In addition, the LECS builds and maintains an ATM address database of LAN Emulation Servers (LES). The LECS maps an ELAN name to a LES ATM address. The LECS performs the following LANE functions:

    • Accepts queries from a LEC
    • Responds to LEC query with an ATM address of the LES for the ELAN/VLAN
    • Serves multiple emulated LANs
    • Manually defined and maintained

    The LECS assigns individual clients to a ELAN by directing them to the LES that corresponds to the ELAN.

        1. LAN Emulation Server (LES)

          LECs are controlled from a central control point called a LAN Emulation Server (LES). LECs communicate with the LES using a Control Direct Virtual Channel Connection (VCC). The Control Direct VCC is used for forwarding registration and control information. The LES uses a Control Distribute VCC, a point-to-multipoint VCC, enabling the LES to forward control information to all the LECs. The LES services the LAN Emulation Address Resolution Protocol (LE_ARP) request which it uses to build an maintain a list of LAN destination MAC addresses.

        2. Broadcast Unknown Server (BUS)

          ATM is based on the notion that the network is point-to-point. Therefore, there is no inherent support for broadcast or any-to-any services. LANE provides this type of support over ATM by centralizing broadcast and multicast functions on a Broadcast And Unknown Server (BUS). Each LEC communicates with the BUS using a Multicast Send VCC. The BUS communicates with all LECs using point-multipoint VCC known as the Multicast Forward VCC. A BUS reassembles received cells on each Multicast Send VCC in sequence to create the complete frame. Once a frame is complete is then sent to all the LECs on a Multicast Forward VCC. This ensures the proper sequence of data between LECs.

        3. LANE Design Considerations

    The following are guidelines for designing LANE services on Cisco routers:

    • The AIP has a bi-directional limit of 60 thousand packets per second (pps).
    • The ATM interface on a Cisco router has the capability of supporting up to 255 subinterfaces.
    • Only one active LECS can support all the ELANs. Other LECS operate in backup mode.
    • Each ELAN has one LES/BUS pair and one or more LECs.
    • LES and BUS must be defined on the same subinterface of the router AIP.
    • Only one LES/BUS pair per ELAN is permitted.
    • Only one active LES/BUS pair per subinterface is allowed.
    • LANE Phase 1 standard does not provide for LES/BUS redundancy.
    • The LECS can reside on a different router than the LES/BUS pair.
    • VCCs are supported over switched virtual circuits (SVCs) or permanent virtual circuits (PVCs).
    • A subinterface supports only one LEC.
    • Protocols such as , AppleTalk, IP and IPX are routable over a LEC if they are defined on the AIP subinterface.
    • AN ELAN should be in only one subnet for IP.
        1. Network Support

    The LANE support in Cisco IOS enables legacy LAN protocols to utilize ATM as the transport mechanism for inter-LAN communications. The following features highlight the Cisco IOS support for LANE:

    • Support for Ethernet-emulated LANs only. There is currently no token-ring LAN emulation support.
    • Support for routing between ELANs using IP, IPX or AppleTalk.
    • Support for bridging between ELANs
    • Support for bridging between ELANs and LANs
    • LANE server redundancy support through simple server redundancy protocol (SSRP)
    • IP gateway redundancy support using hot standby routing protocol (HSRP)
    • DECnet, Banyan VINES, and XNS routed protocols
        1. Addressing

          LANE requires MAC addressing for every client. LANE clients defined on the same interface or subinterface automatically have the same MAC address. This MAC address is used as the end system identifier (ESI) value of the ATM address. Though the MAC address is duplicated the resulting ATM address representing each LANE client is unique. All ATM addresses must be unique for proper ATM operations. Each LANE services component has an ATM address unique form all other ATM addresses.

        2. LANE ATM Addresses

    LANE uses the NSAP ATM address syntax however it is not a Layer 3 network address. The address format used by LANE is :

    • A 13-byte prefix that includes the following fields defined by the ATM Forum:
    • AFI (Authority and Format Identifier) field (1 byte)
    • DCC (Data Country Code) or ICD (International Code Designator) field (2 bytes)
    • DFI field (Domain Specific Part Format Identifier) (1 byte)
    • Administrative Authority field (3 bytes)
    • Reserved field (2 bytes)
    • Routing Domain field (2 bytes)
    • Area field (2 bytes)
    • A 6-byte end-system identifier (ESI)
    • A 1-byte selector field
        1. Cisco's Method of Automatically Assigning ATM Addresses

    The Cisco IOS supports an automated function of defining ATM and MAC addresses. Theses addresses are used in the LECS database. The automation process uses a pool of eight MAC address that are assigned to each router ATM interface. The Cisco IOS applies the addresses to the LANE components using the following methodology:

    • All LANE components on the router use the same prefix value. The prefix value identifies a switch and must be defined within the switch.
    • The first address in the MAC address pool becomes the ESI field value for every LANE client on the interface.
    • The second address in the MAC address pool becomes the ESI field value for every LANE server on the interface.
    • The third address in the MAC address pool becomes the ESI field value for the LANE broadcast-and-unknown server on the interface.
    • The fourth address in the MAC address pool becomes the ESI field value for the LANE configuration server on the interface.
    • The selector field for the LANE configuration server is set to a 0 value. All other components use the subinterface number of interface to which they are defined as the selector field.

    The requirement that the LANE components be defined on different subinterfaces of an ATM interface results in a unique ATM address due to the use of the selector field value being set to the subinterface number.

        1. Using ATM Address Templates

          ATM address definitions is greatly simplified through the use of address templates. However, these templates are not supported for the E.164 ATM address format. The address templates used for LANE ATM addressing can use either an asterisk (*) or an ellipsis (å) character. An asterisk is used for matching any single character. An ellipsis is used for matching leading or trailing characters. Table 6.1 lists the address template value determination.

           

          Unspecified Digits In

          Resulting Value Is

          Prefix (first 13 bytes)

          Obtained from ATM switch via Interim Local Management Interface (ILMI)

          ESI (next 6 bytes)

          Filled using the first MAC address of the MAC address pool plus

           

          0-LANE client

           

          1-LANE server

           

          2-LANE broadcast-and-unknown server

           

          3-LANE Configuration server

          Selector field (last 1 byte)

          Subinterface number, in the range 0 through 255.

          The ATM address templates can be either a prefix, or ESI template. When using a prefix template, the first 13 bytes match the defined prefix for the switch but uses wildcards for the ESI and selector fields. An ESI template matches the ESI field but uses wildcards for the prefix and selector fields.

        2. Rules for Assigning Components to Interfaces and Subinterfaces

    The LANE components can be assigned to the primary ATM interface as well as the subinterfaces. The following are gudielines for applying LANE components on a Cisco router ATM interface.

    • The LECS always runs on the primary interface.
    • Assignment a component to the primary interface falls through to assigning that component on the 0 subinterface.
    • The LES and LEC of the same emulated LAN can be configured on the same subinterface in a router.
    • LECs of two different emulated LANs must be defined on a different subinterface in a router.
    • LESs of two different emulated LANs must be defined on a different subinterface in a router.
        1. Redundancy in LANE environments

    The ATM LANE V 1.0 specification does not provide for redundancy of the LANE components. High avialbility is always a goal for network designers and the single point of failure in the LANE specification requires a technique for redundancy. Cisco IOS supports LANE redundancy through the implmenentation of Simple Server Replicatoin Protocol (SSRP).

    SSRP supports redundancy for LECS and LES/BUS services. LECS redundancy is provided by configuring multiple LECS address in the ATM switches. Each defined LECS is defined with a rank. The rank is the index (number of the entry in the LECS address table) of the LECS address in the table. At iitialization the LECS requests the LECS address table form the ATM swixth. The requesting LECs onreceipt of the LECS addres table tries to connect to all the LECSs with a lower rank. In this way the LECS learns of its role in the redundancy hierarchy. A LECS that connects with a LECS whose rank is higher places itself in a backup mode. The LECS that connects to all other LECS and does not find a ranking higher than its own assumes the responsibility of the primary LECS. In this hierarchy, as shown in Figure 6.2, the failure of a primary LECS does not result in a LANE failure. Rather , the second highest ranking LECS assumes the primary LECS role. Loss of the VCC between the primary and highest ranking secondary signals the highest secondary ranking LECS that it is now the primary LECS.

    In theory any number of LECS can be designed using SSRP. However, Cisco recommends that no more than three LECS be designed into SSRP. The recommendation is based on adding a degree of complexity to the network design which can lead to an increase in the time it takes for resolving problems.

    LES/BUS redundancy using SSRP is similar in that it uses a primary-secondary hierarchy however, the primary LES/BUS pair is assigned by the primary LECS. The LECS determines the primary LES/BUS pair by determining the LES/BUS pair having the highest priority with an open VCC to the primary LECS. The LES/BUS pair priority is assigned during configuration into the LECS database.

    The following guidelines are highly recommended for desinging the LECS redundancy scheme and ensuring a properly running SSRP configuration:

    • Each LECS must maintain the same ELAN database.
    • Configure the LECS addresses in the LECS address table in the same order on each ATM switch in the network.
    • Do not define two LECSs on the same ATM switch when using the Well Known Address. Only one of the LECS will register the Well Known Address with the switch which may led to initialization problems.

    A second type of redundancy mechanism used in LANE is specific to ELANS using IP protocol. The Host Standby Router Protocol (HSRP) enables two routers to share a common virtual IP address using a virtual MAC address assigned to the resulting virtual interface. This enables two routers to respond as the single IP gateway address for IP end stations. Figure 6.3 illustrates the use of HSRP with LANE. The primary and secondary router interface is determined by definition of HSRP on interface or subinterface. HSRP exchanges definition information between the two routers to determine which interface is the primary gateway address. The secondary then sends HELLO messages to the primary to determine its viability. When the secondary does not receive a HELLO message from the primary HSRP router it assumes the primary role.

      1. Data Exchange Interface (DXI)

        ATM networks connect to serial attached routers by implementing the ATM data exchange interface (DXI) specification. The DXI specification enables ATM user-network interface (UNI) connectivity between a Cisco router with only a serial interface to the ATM network. This is accomplished using an ATM Data Service Unit (ADSU). As shown in Figure 6.4, router R1 connects to the ADSU using a High Speed Serial Interface (HSSI) connection. The ADSU recevies data from the router in the ATM DXI format. The ADSU then converts the data into ATM cells and forwards them to the ATM network. The ADSU performs the opposite function for data going to the router.

        1. Supported Modes

    While there are three modes of ATM DXI the Cisco IOS supports only mode 1a. The three modes are:

    • Mode 1a-Supports AAL5 only, a 9232 octet maximum, a 16-bit FCS, up to 1023 virtual circuits.
    • Mode 1b-Supports AAL3/4 and AAL5, a 9224 octet maximum, a 16-bit FCS. AAL5 support up to 1023 virtual circuits. AAL3/4 is supported on one virtual circuit.
    • Mode 2-Supports AAL3/4 and AAL5 with 16,777,215 virtual circuits, a 65535 octet maximum, and 32-bit FCS.
        1. DXI Addressing

    The DXI addressing using a value which is equivalent to a frame relay data link connection identifier. In DXI this field is called a DFA. The ADSU maps the DFA to the appropriate ATM Virtual Path Identifier (VPI) and Virtual Connection Identifier (VCI). Figure 6.5 illustrates the bytes and position mapping of the DXI DFA address to the ATM cell VPI and VCI values.

      1. Classical IP

        Cisco routers are configurable as both an IP client and IP server in support of Classical IP. Classical IP enables the routers to view the ATM network as a Logical IP Subnet (LIS). Configuring the routers as an ATM ARP server enables classical IP networks to communicate over an ATM network. The benefit to this is a simplified configuration. Classical IP support using an ATM ARP server alleviates the need to define the IP network address and ATM address of each end device connecting through the router in the router configuration.

        ATM uses PVCs and SVCs. The ATM ARP server feature of Classical IP is specific to using SVCs. Using the ATM ARP server feature each end device only configures its own ATM address and the address of the ATM ARP server. Since RFC 1577 allows for only one ATM ARP server address there is no redundancy available for Classical IP. As shown in Figure 6.6, the ATM ARP server address can point to a Cisco router. IP clients using Classical IP make a connection to the ATM ARP server address defined in their configuration. The server then sends an ATM Inverse ARP (InARP) request to the client. The client responds with its IP network address and ATM address. The ATM ARP server places these addresses in its cache. The cache is used to resolve ATM ARP requests from IP clients. The IP client established a connection to the IP-ATM address provided in the ATM ARP server reply.

      2. Multiprotocol over ATM (MPOA)

    MPOA provides a single solution for transporting all protocols through an ATM network. MPOA V1.0 in concert with LANE User-to-Network Interface (UNI) V2.0 allows routers and other ATM networking devices to fully exploit VLANs, QoS and high-availability. These network enhancements enable designers to add services while relieving traffic congestion and flexibility to the network. The key benefits to MPOA are:

      • Inter-VLAN "cut-through" which maximizes bandwidth and network segmentation.
      • Robust Layer 3 QoS features to support packetized traffic such as video or voice, while ensuring data service levels.
      • Software only upgrade which minimizes the cost and simplifies implementation.

    The MPOA specification is built on four components. These components are:

      • MPOA Client (MPC)
      • MPOA Server (MPS)
      • Next Hop Resolution Protocol (NHRP)
      • LAN Emulation (LANE)

    Both MPC and MPS functions are supported on Cisco routers. MPOA uses a direct virtual channel connection (VCC) between the ingress (inbound) and egress (outbound) edge or host device. Direct VCCs are also termed shortcut VCCs. The direct VCC enables the forwarding of Layer-3 packets, normally routed through intermediate routers, between source and destination host thereby increasing performance and reducing latency.

    Figure 6.7, illustrates the use of MCP, MPS, and NHRP for establishing a direct VCC between two edge devices servicing two end stations.

        1. Multiprotocol Client (MPC)

    Typically, the Multiprotocol client (MPC) will reside on an ATM edge device such as a Cisco Catalyst family of switches. However, a Cisco router can perform the functions of an MPC or MPS. An MPC provides the following functions:

      • Ingress/egress cache management
      • ATM data-plane and control-plane VCC management
      • MPOA frame processing
      • MPOA protocol and flow detection
      • Identifies packets sent to an MPOA-capable router
      • Attempts to establish a direct VCC with the egress MPC.
        1. Multiprotocol Server (MPS)

    The Multiprotocol server (MPS) provides the forwarding information used by the MPCs. The MPS maintains the information by using Next Hop Resolution Protocol (NHRP). MPS interacts with the NHRP module running in the router. MPS interacts with NHRP in the following manner:

      1. The MPS converts the MPOA resolution request to a NHRP request. The MPS then sends the NHRP request to either the next hop MPS or the Next Hop server (NHS) based on the results form the next hop information search through the MPS tables. MPS ensures that the correct encapsulation is used depending on the next hop server type.
      2. If the next hop is determined to be on a LANE cloud the NHS sends resolution requests to the MPS. Likewise, the NHS sends resolution requests when the destination of the packet is unknown. The MPS may also request the NHS to terminate the request or discard the packet.
      3. If the replies terminate in the router or the next hop interface uses LANE, resolution replies are sent from the NHS to the MPS.
      4. Upon receiving resolution replies from the NHS the MPS sends a MPOA resolution reply to the MPC.

    MPS uses a network ID. The default nework ID for all MPSs is 1. Using different network IDs allows the network designer to segregate traffic. This enables the designer to permit direct VCCs between groups of LECs and deny direct VCCs between others. The network ID of an MPS and NHRP on the same router must be the same in order for reqeusts, replies and shortcuts across the MPS and NHRP.

        1. MPOA Guidelines

    The following is a list of guidelines for designing MPOA:

      • An ELAN identifier must be defined for each ELAN.
      • An MPC/MPS can serve as a single LEC or multiple LECs.
      • A LEC can associate with any MPC/MPS.
      • A LEC can attach to only one MPC and one MPS at a time.
      • A LEC must break its attachment to the current MPC or MPS before attaching another MPC or MPS.
      • A primary ATM interface can have multiple MPCs or MPSs defined with different control ATM addresses.
      • Multiple MPCs or MPSs can be attached to the same interface.
      • The interface attached to the MPC or MPS must be reachable through the ATM network by all LECs that bind to it.
      1. Bandwidth support on routers

    ATM is supported on the Cisco 7500 and 7000 series routers using the ATM Interface Processor (AIP). In designing the ATM internetwork in support of LANE the total ATM bandwidth support for the entire router should not exceed 200 Mbps in full duplex mode. This results in the following possible hardware configurations:

    • Two Transparent Asynchronous Transmitter/Receiver Interface (TAXI) connections.
    • One OC-3 Synchronous Optical Network (SONET) and one E3 connections.
    • One OC-3 SONET and one low-use OC-3 SONET connections
    • Five E3 connections
      1. Configurable Traffic Parameters

    The AIP provides the ability to shape various traffic. The AIP supports up to eight rate queues. Each queue is programmed for a different peak rate. The ATM virtual circuits can be assigned to one of the eight rate queues. A virtual circuit can have an average rate and a burst size defined. The AIP supports the following configurable traffic rate parameters:

      • Forward peak cell rate
      • Backward peak cell rate
      • Forward sustainable cell rate
      • Backward sustainable cell rate
      • Forward maximum burst
      • Backward maximum burst

    Previous | Content