|
|
The Catalyst 3900 and the Catalyst 5000 have ATM expansion modules that provide high-speed connectivity between the switch and an ATM backbone network.
This chapter provides the following information:
ATM is a cell-switching and multiplexing technology that combines the benefits of circuit switching (constant transmission delay and guaranteed capacity) with those of packet switching (flexibility and efficiency for intermittent traffic). Like X.25 and Frame Relay, ATM defines the interface between the user equipment (such was workstations and routers) and the network (referred to as the User-Network Interface [UNI]). This definition supports the use of ATM switches (and ATM switching techniques) within both public and private networks.
Because it is an asynchronous mechanism, ATM differs from synchronous transfer mode methods, where time-division multiplexing (TDM) techniques are employed to preassign users to time slots. ATM time slots are made available on demand, with information identifying the source of the transmission contained in the header of each ATM cell. TDM is inefficient relative to ATM because if a station has nothing to transmit when its time slot comes up, that time slot is wasted. The reverse situation, where one station has large amounts of information to transmit, is also less efficient. In this case, that station can only transmit when its turn comes up, even though all the other time slots are empty. With ATM, a station can send cells whenever necessary.
Another critical ATM design characteristic is its star topology. The ATM switch acts as a hub in the ATM network, with all devices attached directly. This provides all the traditional benefits of star-topology networks, including easier troubleshooting and support for network configuration changes and additions.
Furthermore, ATM's switching fabric provides additive bandwidth. As long as the switch can handle the aggregate cell transfer rate, additional connections to the switch can be made. The total bandwidth of the system increases accordingly. If a switch can pass cells among all its interfaces at the full rate of all interfaces, it is described as nonblocking. For example, an ATM switch with 16 ports set at 155 Mbps would require about 2.5 Gbps aggregate throughput to be nonblocking.
ATM switches transmit data in small units called cells. The latency in a cell switch is very small because of the short cell size. Short cells have a tiny store-and-forward delay. In the absence of port contention and buffering, cells are switched quickly in hardware. For information about the format of an ATM cell, see the "Frame Formats"appendix.
In addition to the low latency, ATM is beneficial to large networks because it:
A PVC is a non-switched connection that is established beforehand (manually pre-provisioned) to satisfy a standing need for network services. It is a logical (not a physical) connection between two communicating ATM peers. This type of connection is typically established by a network administrator.
User applications that require an on-going, specific level of transmission bandwidth typically use PVCs for interconnectivity. The network bandwidth required in this type of application tends to be more predictable and constant, enabling the physical transmission medium to be tailored to an expected volume of traffic, and vice versa.
With a PVC, everything is statically configured and no signaling is involved. The PVC is mapped to a network in a subinterface point-to-point configuration. The logical data link layer can use SNAP encapsulation (as defined in RFC 1483). This allows multiple protocols to be multiplexed over one PVC. Alternatively, the logical data link layer can use LANE Version 1 over PVC.
The PVC is statically mapped at each ATM node. The path of the PVC is identified at each switch by an incoming virtual channel identifier/virtual path identifier (VCI/VPI) and an outgoing VCI/VPI.
An SVC is a switched connection that is established by a defined and standardized ATM signaling protocol. This type of connection is set up dynamically (on demand) across the network, as required by the user's communications applications. An SVC is established and torn down using a flexible connection setup protocol that supports various connection types.
The transfer of information between network users by means of SVCs typically occurs through shared network facilities, rather than through dedicated transmission lines or owned physical facilities.
Establishing an ATM SVC involves an agreement between the end nodes and all the switches in between. Each end node has a special signaling channel to the connected switch called the UNI. Switches have a signaling channel between them called the Network-to-Network Interface (NNI). Cells that arrive on the signaling channel are reassembled into frames in the reliable Service Specific Connection Oriented Protocol (SSCOP). The signaling information follows the Q.2931 standard.
Establishing an SVC potentially involves signaling between the following:
The UNI is defined by the ATM Forum UNI specification.
Interfaces to public ATM networks are identified by an E.164 address. Interfaces to private ATM networks are identified by a network service access point (NSAP) address. These addresses are contained in different fields of the same 20-octet address.
Once an SVC is established, it functions like a PVC. SVCs can be used in point-to-point subinterface configuration or point-to-multipoint nonbroadcast multiaccess (NBMA) configuration.
The purpose of the ATM adaptation layer (AAL) is to receive the data from the various sources or applications and convert or adapt it to 48-byte segments that will fit into the payload of an ATM cell. The AAL segments upper-layer user information into ATM cells at the transmitting end of a virtual connection and reassembles the cells into a user-compatible format at the receiving end of the connection. These complimentary functions occur between communicating peers in the network at the same level of the ATM architectural model.
The AAL is not a network process. Rather, AAL functions are performed by the user's network terminating equipment on the user side of the UNI. Consequently, the AAL frees the network from concerns about different traffic types.
How AAL processes are carried out depends on the type of traffic to be transmitted. Different types of AALs handle different types of traffic, but all traffic is ultimately packaged by the AAL into 48-byte segments for placement into ATM cell payloads. Consequently, several different AALs have been defined for different types of services.
Table 4-1 lists these AALs
| Traffic Class | Timing Relationship | Connection Mode | Bit Rate | Traffic Description |
|---|---|---|---|---|
Class A (AAL1) | Synchronous | Connection- oriented | Constant | This type of traffic typically consists of constant bit rate (CBR) analog signals. Hence, synchronous timing relationships exist between the senders and receivers of this traffic. This type of traffic over an ATM network is often referred to as circuit emulation service, an example of which is fixed bit rate, uncompressed voice, or video data. |
Class B (AAL2) | Synchronous | Connection- oriented | Variable | As with Class A traffic, synchronous timing relationships exist between the senders and receivers of Class B traffic. However, Class B relates to variable bit rate (VBR) traffic, typical examples of which are compressed voice and video traffic. Such traffic is typically "bursty" in nature. |
Class C (AAL3/4) | Asynchronous | Connection- oriented | Variable | No timing relationships exist between the senders and receivers of data. Hence, such traffic is asynchronous. Class C handles VBR connection-oriented traffic. This class provides point-to-point or point-to-multipoint ATM cell relay services over connections established "on the fly" through signaling and routing messages exchanged between data senders and receivers. This service handles multiple traffic types (data, voice, and video) in which user data is arranged into ATM cells for efficient transport through the network. Class C of traffic contains sequencing bits that allows the cells to take different paths and still be reassembled in the correct order at the receiving station.
This type of traffic is sensitive to cell loss, but not to cell transport delay (or latency). Latency is the delay between the time a device receives a cell on its input port and the time the cell is forwarded through its output port. |
Class D (AAL5) | Asynchronous | Connectionless | Variable | Class D handles unspecified bit rate (UBR) traffic in a connectionless, asynchronous manner. |
Because ATM is inherently a connection-oriented transport mechanism and because the current applications of ATM are heavily oriented toward LAN traffic, many of the current ATM products, including the Catalyst 3900 and the Catalyst 5000, support the Class D adaptation layer with AAL5.
The AAL performs two main functions in service-specific sublayers of the AAL:
AAL5 has been designed to process traffic typical of today's LANs. Originally, AAL3/4 was designed to process this kind of traffic. However, the inefficiency of AAL3/4 for handling LAN traffic led to the use of AAL5 for such traffic.
AAL5 provides a streamlined data transport service that functions with less overhead than AAL3/4. AAL5 is typically associated with UBR traffic.
Another AAL5 attribute contributing to its efficiency is that is uses only 5 bytes of header. None of the payload is used for header information. Also, AAL5 calculates a 32-bit cyclic redundancy check (CRC) over the entire AAL5 protocol data unit to detect cell loss and the incorrect ordering or incorrect insertion of cells.
For purposes of AAL5 traffic processing, the CS is divided into the following parts:
The building blocks of an ATM internetwork may consist of the following:
The Catalyst 3900 ATM expansion module supports up to 63 VLANs (or ELANs). Each ELAN corresponds to a TrCRF. Each association between the ATM expansion module and a TrCRF creates a virtual ATM port. A virtual ATM port is the equivalent of an LAN Emulation Client (LEC).
LANs can use connectionless service. However, ATM is always a connection-oriented service. Devices first use a signaling process to establish a path with an ATM destination. Devices can send cell-based traffic only after the devices have identifiers pointing to the connection path.
LANE uses point-to-multipoint connections to service the connectionless broadcast service that is required by LAN protocols.
Cisco's Token Ring implementation of LANE makes an ATM interface look like one or more Token Ring interfaces. Setting up LECs allows the Catalyst 3900 or Catalyst 5000 Token Ring module to operate in an ATM LAN environment containing Cisco 7000 or Cisco 4500 series routers with ATM Interface Processor (AIP) connected to a LightStream 1010 ATM switch.
Figure 4-1 illustrates the physical layout of an ATM network that uses LANE.
Figure 4-2 illustrates the logical view of the LANE network.
LANE is an ATM service defined by the ATM Forum specification LAN Emulation over ATM (ATM_FORUM 94-0035). This service emulates the following LAN-specific characteristics:
LANE service provides connectivity between ATM-attached devices and LAN-attached devices. This includes connectivity between ATM-attached stations and LAN-attached stations as well as connectivity between LAN-attached stations across an ATM network.
Because LANE connectivity is defined at the MAC layer, upper protocol-layer functions of LAN applications can continue unchanged when the devices join ELANs. This feature protects corporate investments in legacy LAN applications.
An ATM network can support multiple independent ELANs. Membership of an end system in any of the ELANs is independent of the physical location of the end system. The end systems can move easily from one ELAN to another, regardless of whether or not the hardware is moved.
A Catalyst 3900 or Catalyst 5000 ATM module can participate in up to 63 of these ELANs.
LANE is defined on a client-server LAN model, as follows:
The Catalyst 3900 ATM module currently supports only the LEC function. A Catalyst 5000 or a Cisco 7000, Cisco 7200, Cisco 7500, RSP 7000, Cisco 4500, or Cisco 4700 with an AIP can supply all LANE functions.
Communication among LANE components is typically handled by several types of VCCs. Some VCCs are unidirectional; others are bidirectional. Some are point-to-point and others are point-to-multipoint. Figure 4-3 illustrates the various types of VCCs.
The following section describes the processes involved with a client requesting to join an ELAN.
The following process (illustrated in Figure 4-3) normally occurs after an LEC has been enabled on the ATM module:
Step 1. The client requests to join an ELAN. The client sets up a connection to the LECS to find the ATM address of the LANE server for its ELAN. See the bidirectional, point-to-point link (link 1-7 in Figure 4-3).
An LEC finds the LECS using the following methods in the listed order:
Step 2. The LECS identifies the LES. Using the same VCC, the LECS returns the ATM address and the name of the LES for the client's ELAN.
Step 3. The client tears down the configure direct VCC.
Step 4. The client contacts the server for its LAN. The client sets up a connection to the LES for its ELAN (bidirectional, point-to-point control direct VCC [link 1-7 in Figure 4-3]) to exchange control traffic. Once a control direct VCC is established between an LEC and LES, it remains up.
Step 5. The LES verifies that the client is allowed to join the ELAN. The server for the ELAN sets up a connection to the LECS to verify that the client is allowed to join the ELAN (bidirectional, point-to-point server configure VCC [link 11-12 in Figure 4-3]).
The server's configuration request contains the client's MAC address, its ATM address, and the name of the ELAN. The LECS checks its database to determine whether the client can join that LAN; then it uses the same VCC to inform the server whether or not the client is allowed to join.
Step 6. The LES allows or disallows the client to join the ELAN. If allowed, the LES adds the LEC to the unidirectional, point-to-multipoint control distribute VCC (link 2-8 in Figure 4-3) and confirms the join over the bidirectional, point-to-point control direct VCC (link 1-7 in Figure 4-3). If disallowed, the LES rejects the join over the bidirectional, point-to-point control direct VCC (link 1-7 in Figure 4-3).
Step 7. The LEC sends LE_ARP packets for the broadcast address, which is all 1s. Sending LE_ARP packets for the broadcast address returns the ATM address of the BUS. Then the client sets up the multicast send VCC (link 4-9 in Figure 4-3) and the BUS adds the client to the multicast forward VCC (link 5-10 in Figure 4-3) to and from the BUS.
Step 8. The LEC registers the ring numbers of all other TrCRFs within its TrBRF that contain active ports on the local switch.
On a LAN, packets are addressed by the MAC-layer address of the destination and the source stations. To provide similar functionality for LANE, MAC-layer addressing must be supported. Every LEC must have a MAC address. In addition, every LANE component (server, client, BUS, and configuration server) must have a unique ATM address.
All LECs on the same interface have a different, automatically assigned MAC address. That MAC address is also used as the end-system identifier part of the ATM address as explained in the following sections.
A LANE ATM address has the same syntax as an NSAP, but it is not a network-level address. It consists of the following:
The Catalyst 3900 and Catalyst 5000 ATM modules use ILMI registration to build their ATM addresses and to register the addresses with the ATM switch. To build its ATM address, each module obtains its ATM address prefix from the ATM switch. Then it combines the ATM address prefix with its own MAC address and the selector value of 0 (zero). Once the ATM module has determined its ATM address, it uses ILMI to register this address with the ATM switch.
As communication occurs on the ELAN, each client dynamically builds a local LE_ARP table. The LE_ARP table maps ELAN MAC addresses (Layer 2) to ATM addresses (also Layer 2). A client's LE_ARP table can also have static, preconfigured entries. The LE_ARP table maps MAC addresses to ATM addresses.
When a client first joins an ELAN, its LE_ARP table has no dynamic entries and the client has no information about destinations on or beyond its ELAN.
To learn about a destination when a packet is to be sent, the client begins the following process to find the ATM address corresponding to the known MAC address:
Step 1. The client sends an LE_ARP request to the LANE server for this ELAN (point-to-point control direct VCC [link 1-7 in Figure 4-3]).
Step 2. If the MAC address is registered with the server, it returns the corresponding ATM address. If not, the LES forwards the LE_ARP request to all clients on the ELAN (point-to-multipoint control distribute VCC [link 2-8 in Figure 4-3]).
Step 3. Any client that recognizes the MAC address responds with its ATM address (point-to-point control direct VCC [link 1-7 in Figure 4-3]).
Step 4. The LES forwards the response (point-to-multipoint control distribute VCC [link 2-8 in Figure 4-3]).
Step 5. The client adds the MAC address-ATM address pair to its LE_ARP cache.
Step 6. Now the client can establish a VCC to the desired destination and proceed to transmit packets to that ATM address (bidirectional, point-to-point data direct VCC [link 6-6 in Figure 4-3]).
For unknown destinations, the client sends a packet to the BUS, which forwards the packet to all clients. The BUS floods the packet because the destination might be behind a bridge that has not yet learned this particular address.
The Catalyst 3900 allows you to define up to 64 traffic profiles. These profiles can be used to define the maximum rates for each traffic type.
For each VLAN (or ELAN), a traffic profile must be mapped to each DD-VCC. The process of mapping depends on whether the traffic is incoming or outgoing:
When an LEC has broadcast or multicast traffic, or unicast traffic with an unknown address to send, the following process occurs:
Step 1. The client sends the packet to the BUS (unidirectional, point-to-point multicast send VCC [link 4-9 in Figure 4-3]).
Step 2. The BUS forwards (floods) the packet to all clients (unidirectional, point-to-multipoint multicast forward VCC [link 5-10 in Figure 4-3]).
This VCC branches at each ATM switch. The switch forwards these packets to multiple outputs. (The switch does not examine the MAC addresses; it simply forwards all packets it receives.)
This section describes some scenarios for using the Catalyst Token Ring switches with the ATM expansion module. Figure 4-4 shows how the ATM expansion module can be used to connect Catalyst Token Ring switches through an ATM switch.
The ATM expansion module is well suited for the following environments:
Because one ATM expansion module can support as many LECs as there are VLANs in a Catalyst Token Ring switch, the question arises: Why use two ATM expansion modules?
The first reason is to provide a backup LEC. By enabling two LECs on two different ATM expansion modules to be members of both the same VLAN and the same ELAN, the Catalyst Token Ring switch's spanning-tree operation will automatically use one LEC for forwarding frames and the other LEC for blocking frames (active standby). For the backup LEC configuration to work, spanning tree must be enabled for the related LAN switch domain.
The second reason to use two ATM expansion modules is to increase system resources. In some environments the resources associated with a single ATM expansion module might affect individual LEC performance. By moving one or more LECs to a second ATM expansion module, you can significantly increase the resources available to each LEC.
Before installing and configuring the ATM module, be aware of the following:
Posted: Wed Oct 2 03:45:50 PDT 2002
All contents are Copyright © 1992--2002 Cisco Systems, Inc. All rights reserved.
Important Notices and Privacy Statement.