MPOA Configuration

MPOA Configuration

Surprisingly, in spite of all of the background complexity of MPOA, configuring the MPS and MPC is quite simple. You must first have LANE configured, though. Without proper LANE configurations, MPOA never works.

Note
Before attempting to configure MPOA on your Catalyst LANE module, ensure that you have an MPOA-capable module. The legacy LANE modules do not support MPOA. The MPOA-capable modules include hardware enhancements to support the MPC functions.

Generally, the LANE module must be a model number WS-X5161, WS-X5162 for OC-12, or WS-X5167 and WS-X5168 for OC-3c support.

Although you can configure the MPOA components in any sequence, the following sequence helps to ensure that the initialization processes acquire all necessary values to enable the components.

  1. Configure LECS database with elan-id
  2. Enable LECS
  3. Configure MPS
  4. Configure MPC
  5. Enable LANE server and bus
  6. Enable LANE clients

A configuration sequence other than that listed does not prevent MPOA from functioning, but you might need to restart the LANE components so that MPOA can correctly operate. Specifically, you should ensure that the LEC can acquire the elan-id from the LECs before you enable the LEC. The elan-id is used by the MPOA components to identify broadcast domain membership. This is useful when establishing shortcuts.

Cisco implementations of LANE and MPOA use a default NSAP address scheme. Chapter 9 describes the NSAP format in detail. Remember, however, that the NSAP is comprised of three parts:

  • The 13-byte prefix
  • The 6-byte end-station identifier (esi)
  • The 1-byte selector (sel)

The show lane default command enables you to see what the NSAP address for each LANE component will be if you enable that service within that device. Cisco’s implementation of MPOA also uses a default addressing scheme that can be observed with the show mpoa default command. Example 10-1 shows the output from these two commands.

Example 10-1 LANE and MPOA Component Addresses

Note that the esi portion highlighted in italics of the MPS and MPC NSAP continue to increment beyond the esi portion of the LECS NSAP address. The selector byte, however, does not correlate to a subinterface as happens with the LANE components. Rather, the selector byte indicates which MPS or MPC sources the traffic. A host, edge device, or router can have more than one MPC or MPS enabled. The selector byte identifies the intended device.

Configuring the LECS Database with elan-id

In addition to the LECS configuration statements necessary to enable LANE, another database statement must be present to enable the MPOA servers and clients to identify their membership in a broadcast domain. Each broadcast domain (ELAN) in the ATM domain must be uniquely identified with a numerical ELAN identifier value. The elan-id value is a 4-octet value. Every LEC in an ELAN must have the same elan-id. Every ELAN serviced by the LECS must have a unique elan-id. Because the MPC and MPS must associate with a LEC, the MPC and MPS have an elan-id by association. They have the elan-id of the LEC.

The syntax to identify the ELAN is as follows:

Example 10-2 shows how to configure the LECS database for MPOA.

Example 10-2 LECS Database Configured for MPOA

Every ELAN with MPC components must have an elan-id assigned to it. In Example 10-2, two ELANs are defined, elan1 and elan2, each with a unique elan-id of 101 and 102, respectively. The actual value used does not matter, so long as the value is unique to the ATM domain.

Rather than letting the LECs obtain the elan-id value from the LECS, you can manually configure the elan-id value in each of the LECs in your network. But this can become administratively burdensome and is not, therefore, a widely used approach. By having the elan-id configured in the LECS database, you simplify your configuration requirements by placing the value in one location rather than many.

Whenever a LANE client connects to the LECS as part of the initialization process, the LEC acquires the elan-id. This value is then used by the MPS and the MPC during their initialization processes, and during the shortcut establishment.

Confirm that the LEC acquired the elan-id with the show lane client command. The bold highlight in Example 10-3 indicates that the Cat-A LEC belongs to the ELAN with an elan-id value of 101. This value came from the ELAN configuration statement in the LECS database.

Example 10-3 show lane client with ELAN-ID Acquired from LECS

If you enable the LANE components before you put the necessary MPOA statements in the LECS database, the LANE components do not acquire the elan-id value. In this case, you need to restart the LEC so it reinitializes and obtains the elan-id. Without the elan-id, the MPS and MPC cannot establish neighbor relationships. Nor can the egress MPS issue a cache imposition request, as the elan-id is one of the parameters passed in the request as defined by MPOA.

Configuring the MPS

Configuring the MPS requires three categories of configuration:

  • Global configurations to set MPS parameters
  • Major interface configurations to enable the server
  • Subinterface configurations to associate LECs with the MPS

To create a functional MPOA system, you must have an MPS at the ingress and egress points in the network to resolve MPC to MPC values.

MPSs must be created on routers to interact with NHSs and routing tables. In the Catalyst product line, you can enable an MPS on a route-switch module (RSM) if the RSM has a versatile Interface Processor 2 (VIP2) containing an ATM port adapter. You cannot enable an MPS on the LANE module.

MPS Global Configuration

You have the option to modify MPS default configuration parameters. For example, you can modify the MPS’s control NSAP address to a value other than the one used by default Cisco methods.

MPOA defines several timers and default values for those timers. You can modify the defaults if you so choose. Table 10-1 shows the names and descriptions for each of the MPS configurable timers.

Table 10-1. MPS Configurable Timers

Timer NameTimer Description
Keepalive-timeHow often the MPS issues a keepalive frame to the MPC. The default value is 10 seconds. Values can range from 1 second to 300 seconds.
Keepalive-lifetimeHow long the MPC should consider a keepalive valid. This should be at least three times the keepalive-time value. The default value is 35 seconds. Values can range from 3 seconds to 1000 seconds.
Holding-timeHow long an MPC should retain an MPOA resolution reply. The default is 20 minutes (1200 seconds). Values can range from 1 minute to 120 minutes.

The MPS issues keepalive messages to the neighbor MPCs at a frequency defined by the keepalive-time value. This maintains the neighbor relationship between the MPC and the MPS.

Example 10-4 shows a sample global configuration for the MPS. The global configuration requires you to name the MPS. The name is only locally significant, so you can name it whatever you want. If you enable more than one MPS in the unit, they need to be uniquely named within the device.

Example 10-4 MPS Global Configuration Example

The network-id parameter allows you to prevent shortcuts between LECs on ELANs served by one MPS and LECs on ELANs served by another MPS. By default, all MPSs belong to network-id 1. If you have two MPSs, one with a network-id of 1 and the other with a network-id of 2, LECs associated with network-id 1 cannot develop a shortcut to LECs associated with network-id 2.

  • Tip
    Even if you elect to retain the default values, you must still enter the global configuration statement mpoa server config name MPS_server_name.

MPS Major Interface Configuration

Enabling the MPS requires one statement on the ATM major interface. Note that, although multiple MPSs can be enabled in a router, even on the same major interface, an MPS can be associated with only one hardware interface. If you have two ATM modules, you need to enable at least two uniquely named MPSs. Example 10-5 shows the configuration required on the ATM major interface. The single statement enables the MPS.

Example 10-5 MPS Major Interface Configuration Example

MPS Subinterface Configuration

The MPS must have an LEC associated with it so the MPS can communicate with other MPOA devices in the ELAN. Although more than one MPS can be associated with an interface, you can bind an LEC to only one MPS. Example 10-6 shows how to associate an MPS with an LEC. Note that only one statement is required to do this. If you enter the lane client mpoa command before you actually enable the LEC, you receive an alert informing you of this. This is all right and can be ignored, as long as you remember to enable an LEC. Without the LEC, the MPS is operating, but incapable of supporting MPOA operations. This results from its inability to communicate with any other MPOA devices in the ELAN.

Example 10-6 MPS Subinterface Configuration Example

Examining the MPOA Server

If you correctly configure the MPS, you should see output as shown in Example 10-7.

Example 10-7 Output from show mpoa server Command

In Example 10-7, the MPS sees two MPC neighbors. The output displays the virtual circuits used to communicate with each of the MPCs. These circuits should not experience idle timeouts and should, therefore, remain open. The default idle timeout for Cisco equipment is 5 minutes. If the device sees no traffic on the circuit for the idle timeout value, it releases the circuit. However, by default, the MPS issues keepalives to MPCs every 10 seconds. You can modify this value, but generally you should leave the timers at the default values.

If any of the neighbors are MPSs, the display would also present their NSAP address along with the MPC addresses.

Configuring the MPC

Like the configuration tasks for the MPS, configuration of the MPC requires three categories of configuration:

  • Global configurations to define MPC parameters
  • Major interface configurations to enable the server
  • Subinterface configurations to associate LECs with the MPC

MPC Global Configuration

From the LANE module global configuration prompt, you can specify MPC parameters to modify the default behavior. Specifically, you can change the MPCs NSAP address from the Cisco default value displayed in the show mpoa default command. You can also modify the threshold that causes the MPC to issue an MPOA resolution request to the MPS. The MPOA document specifies a default threshold of 10 frames in 1 second. If the MPC observes a flow with more than 10 frames per second between devices in two different ELANs, it issues the resolution request. You can modify the frame count and the integration period. Example 10-8 shows a sample global configuration for an MPC in a Catalyst LANE module, along with some results from the Help menu. Multiple MPCs can be enabled in your Catalyst, but each must be uniquely named and individually configured.

Example 10-8 MPC Global Configuration Example

  • Tip
    Even if you elect to retain the default values of 10 frames per second, you must still enter the global configuration statement mpoa client config name MPC_client_name.

MPC Major Interface Configuration

As with the MPS configuration, the MPC is not enabled until you enter the mpoa client command on the major interface. Example 10-9 shows a major interface configuration to enable the client. Notice its similarity to the command to enable an MPS as in Example 10-5.

Example 10-9 MPC Major Interface Configuration Example

MPC Subinterface Configuration

Each VLAN for which you want MPOA capability must have an MPC associated with an LEC on the appropriate subinterfaces. Example 10-10 demonstrates how to build the LEC to MPC association. A three-way relationship exists in the configuration. In Example 10-10, the first lane client command enables the LEC/VLAN to participate in MPOA. The LEC associates with a VLAN 12 in the second lane client command. VLAN 12 and ELAN_name correlate to the same broadcast domain.

Example 10-10 MPC Subinterface Configuration Example

Remember that you can bind the LEC to only one MPC. If you bind the MPC to the LEC before you enable the LEC, you receive a warning indicating that no LEC is configured. You can ignore this warning as long as you remember to eventually enable a LEC. When you complete the lane client mpoa command, you can create the LEC on the subinterface. If you create the LEC first, you can enable the MPC afterwards. In either case, make sure that the LEC acquires the elan-id.

Examining the MPOA Client

If you configure MPOA correctly, you should see output for your client as shown in Example 10-11. This output displays the MPC situation before any shortcuts are established. You can identify shortcuts when there are entries under the Remote Devices known section of the MPC output display.

Example 10-11 MPOA Client Show Output

Suppose that you send an extended ping to an egress MPC with 20 pings in the sequence. When the ingress MPC sends enough frames to cross the shortcut threshold, it issues the shortcut request to its neighbor MPS. Assuming that the MPS resolved the shortcut, the ingress MPC can establish the shortcut and start using it rather than the default path.

But during the extended ping operation, the egress MPC sends an echo reply for each echo request issued from the ingress MPC. If the shortcut threshold is set for the same or less than for the ingress MPC, the echo replies cause the egress MPC to issue a shortcut request too. Ultimately, both the ingress and egress MPCs develop shortcuts between each other as illustrated by Example 10-12.

Example 10-12 show mpoa client Display after Shortcut

The MPCs use only one of the shortcuts, though. The shortcut established by the MPC with lowest NSAP address is used by both clients. The ingress MPC of Example 10-12 has the lowest MAC address and should be the one used by the two MPCs. Issuing the show atm vc 49 command confirms that the local device originated from virtual circuit (VC) 49 (see Example 10-13).

Example 10-13 Confirm Call Origination

VC 47 eventually times out and disappears. Until a device tears down the circuit, both circuits remain in place, but only one is used. Note that the traffic counters for VC 47 show zero frames sent or received. Whether or not this happens is a function of your traffic flows. Not all application/protocol flows create a one for one reply-response and will therefore only create one VC.

About the author

Scott

Leave a Comment