skip book previous and next navigation links
go up to top of book: HP OpenVMS System Manager's Manual, Volume 2:... HP OpenVMS System Manager's Manual, Volume 2:...
go to beginning of chapter: Managing the LAT Software Managing the LAT Software
go to previous page: Understanding the LAT Network Understanding the LAT Network
go to next page: Understanding the LATCP UtilityUnderstanding the LATCP Utility
end of book navigation links

Understanding LAT Configurations  



When you set up a LAT system, you need to understand the relationship between the LAT software and the network so you can configure your system to operate efficiently. The following sections contain information that will help you understand the following concepts:

LAT Relationship to OpenVMS Clusters and DECnet  

Although the LAT protocol works independently of OpenVMS Cluster software, HP recommends that you configure a service node to complement the OpenVMS Cluster concept. You achieve this by creating a service on each node in an OpenVMS Cluster and assigning the cluster name to this service. A terminal server assesses the availability of cluster services and establishes a connection to the node that is least busy. Thus, the LAT protocol helps balance the cluster load. If one node in the cluster fails, the terminal server can transfer the failed connections to another service node within the cluster.

The LAT software does not use DECnet as a message transport facility, but instead uses its own virtual circuit layer to implement a transport mechanism. The LAT and DECnet software work independently in a common LAN environment. For compatibility, if a service node is also a DECnet node, the service node name should be the same as the DECnet node name.

LAT and DECnet Running on the Same Controller  

If Ethernet ports will be running both DECnet and LAT, you must start the DECnet software before the LAT software. If you do not start DECnet software first, all existing LAT connections may terminate, and reconnecting to the system via LAT may not be possible.

LAT and DECnet Running on Different Controllers  

If DECnet is configured on the system (or if the system is part of a cluster), the SCSSYSTEMID system paramater may contain a nonzero value. Normally, this is not a problem unless the system has two or more LAN controllers connected to the same logical LAN.

For example, if your system has an FDDI controller and an Ethernet controller, your site may be configured so that the FDDI ring attached to the FDDI controller and the Ethernet segment attached to the Ethernet controller are bridged by a 10/100 LAN bridge (FDDI-to-Ethernet). In this configuration, it is impossible to run LAT over both controllers.

In such a configuration, you must run LAT and DECnet over the same controller if SCSSYSTEMID is not 0. If they do not run on the same controller, DECnet starts first, which in turn causes the LAT startup on the other controller to fail. This failure occurs because LAT startup tries to use the AA-00-04-00-xx-xx address (the DECnet LAN address); however, because DECnet is already using this address on another controller, the data link layer prevents the LAT startup from using that address. (In a single logical LAN, all data link addresses must be unique. Because both controllers try to use the same address, it is no longer unique.)

Using the following command to create the LAT link also fails because the LAN driver tries to use the address based on SCSSYSTEMID:

LATCP> CREATE LINK LAT$LINK_2 /NODECNET
If SCSSYSTEMID is set to 0, configuring LAT and DECnet on different controllers is possible. However, in a cluster environment, SCSSYSTEMID cannot be set to 0.

Using Multiple LAN Adapters  

When you use multiple LAN addresses for one LAT node, you can configure a system with multiple LAN adapters connected to the same logical LAN. The LAT software can run over each adapter simultaneously and can better maintain connections. For example, when a virtual circuit chooses a primary path and uses it for all LAT message transmissions, the LAT software can continue communications through another adapter or logical path if that original path becomes blocked.


NoteNodes running versions of LAT software prior to Version 5.3 of the LAT protocol (included in the OpenVMS operating system beginning with Version 7.0) may exhibit some differences in behavior. Therefore, if your configuration includes earlier versions of the LAT software, such as Version 5.1 or Version 5.2, note the differences and considerations discussed in this chapter.

Supported Configurations  

Although it is possible to run LAT over multiple LAN adapters, it is still not possible to route LAT from one logical LAN to another. The following examples show supported LAT configurations for nodes running Version 5.3 of the LAT protocol (including nodes running Version 5.2 and 5.1 as well).

The widely used configuration shown in Multiple Address LAT Configuration: One LAN with Mixed-Version LAT Nodes has an OpenVMS system running LAT Version 5.3 software over two Ethernet adapters (labeled A and B in the diagram) connected to the same physical LAN as a DECserver 200.  

Figure 2  Multiple Address LAT Configuration: One LAN with Mixed-Version LAT Nodes  
LAN with mixed-version LAT nodes

When a LAT connection is started between the DECserver 200 and the OpenVMS system, the LAT software determines that it is possible to use both adapters A and B for the LAT virtual circuit. One of the adapters will be chosen as the primary communications path while the other will be present in the unlikely event that the primary path fails.

For example, if a user connects to the OpenVMS system from the DECserver 200, the OpenVMS system determines that there are two paths but chooses adapter B as the primary communications path. If the user runs a program that generates a large amount of output from the OpenVMS system and adapter B fails in some manner during the output, the LAT software will attempt to continue communications from the OpenVMS system to the DECserver through adapter A.

Multiple Address LAT Configuration: Two LANs with Mixed Version LAT Nodes shows two LANs bridged together. However, this configuration will have the same characteristics as the configuration shown in Multiple Address LAT Configuration: One LAN with Mixed-Version LAT Nodes. 

Figure 3  Multiple Address LAT Configuration: Two LANs with Mixed Version LAT Nodes  
multiple address LAT config


NoteIt is possible for Ethernet 2 in Multiple Address LAT Configuration: Two LANs with Mixed Version LAT Nodes to be an FDDI network. The LAT software regards each adapter as a network path with equal point-to-point communications and does not treat FDDI controllers any differently. However, for large buffer support, see Large Buffers in Ethernet/FDDI Configurations for more details.

In Multiple Address LAT Configuration: Two LANs with Version 5.3 LAT Nodes, any virtual circuit created between the two OpenVMS systems will have two paths: through controllers B and C or A and D. If one path fails, the virtual circuit will continue over the other path. If both paths fail, the virtual circuit will eventually time out.  

Figure 4  Multiple Address LAT Configuration: Two LANs with Version 5.3 LAT Nodes  
LAT config example

Unsupported Configuration  

When configuring a network to use an OpenVMS system running the LAT Version 5.3 software, avoid the configuration shown in Unsupported Multiple Address LAT Configuration. 

Figure 5  Unsupported Multiple Address LAT Configuration  
LAT config

Any configuration similar to this diagram will result in unpredictable results and may not function. In a network environment, LAT Version 5.1 and 5.2 nodes can have only a single logical LAN address. The configuration in Unsupported Multiple Address LAT Configuration violates this rule. The configuration shown in Multiple Address LAT Configuration: Two LANs with Version 5.3 LAT Nodes is a valid alternative.

Creating Logical LAT Links  

The LAT software regards all paths as equal, point-to-point communications. The LAT software can support a maximum of eight LAN adapters simultaneously (and it is possible to connect all controllers to the same logical LAN). To get the maximum coverage over possible path failures, each logical link should be created prior to setting the LAT node state to ON in SYS$MANAGER:LAT$SYSTARTUP.COM.

For example, if a system has one Ethernet adapter (device ESA0) with two FDDI adapters (FCA0 and FCB0) and the system manager chooses to run LAT over all adapters, the LAT$SYSTARTUP.COM file would contain the following commands:

	$!
	$! Create each logical LAT link with a unique name and
	$! unique LAN address (forced with /NODECNET).
	$!
	$ LCP CREATE LINK ETHERNET /DEVICE=ESA0 /NODECNET
	$ LCP CREATE LINK FDDI_1 /DEVICE=FCA0 /NODECNET
	$ LCP CREATE LINK FDDI_2 /DEVICE=FCB0 /NODECNET
	$!
	$! Turn on the LAT protocol.
	$!
	$ LCP SET NODE /STATE=ON

CautionIf the LATCP command SET NODE /STATE=ON is entered before the link is created, a random or default LAT$LINK will be created on one of the LAN adapters. There is no way to predict which LAN adapter will be chosen (it is dependent on the system configuration). Therefore, all logical LAT links should be created before LAT is started.

Be sure each logical link is created with the /NODECNET qualifier. It will prevent the possibility of link creation failure if multiple adapters attempt to use the DECnet style address. Having more than one LAN adapter connected to the same logical LAN with the same address violates LAN conventions and will cause problems with LAT and other protocols.


It is possible to create logical LAT datalinks after the LAT protocol has been started. Any existing virtual circuit will attempt to find any new paths through the newly created logical datalink when it is ready for use. However, HP does not recommend that you create links at this point because during the time it takes existing virtual circuits to discover new paths through this newly created datalink, the virtual circuit may fail before the new paths are discovered.

Path Discovery  

The OpenVMS LAT software uses a combination of the directory service and solicitation to obtain paths for each virtual circuit. To expedite path discovery at virtual circuit startup, HP recommends that you configure a system with multiple LAN adapters to maintain a LAT service and node database, by performing the following actions:

An OpenVMS system running with outgoing connections disabled and no service and node database is still capable of running with multiple paths for each virtual circuit. These paths must be discovered through the LAT solicitation process and will take longer (leaving the possibility for virtual circuit failure to occur before all paths have been discovered).

Modifying LAT Parameters  

In the unlikely event of a path failure, it will take the OpenVMS LAT software time (which will vary depending on the number of adapters to which the remote node has access) to locate another working path. Therefore, HP strongly recommends that you modify the following LAT parameters on potential LAT master nodes:

Although it is possible to keep virtual circuits running through multiple adapters to LAT Version 5.1 or LAT Version 5.2 master nodes, there is still a possibility that the connections to these nodes may fail.

LAT Version 5.2 and LAT Version 5.1 master nodes do not have the ability to recognize multiple paths to LAT nodes that provide services. They can only communicate with such nodes through one remote address at a time. Therefore, if a LAN path failure occurs when a LAT master node running LAT Version 5.1 or Version 5.2 attempts to connect to a remote LAT Version 5.3 node providing services, the LAT Version 5.3 node might not discover this failure in time and the LAT master node may time out the connection. You can partially solve this problem by increasing the retransmit limit to as high a setting as possible.

In addition, if a LAT Version 5.3 node providing services views the virtual circuit as completely idle during the primary path failure, no attempt will be made to use any of the alternate paths (because of the previously described LAT Version 5.2 and 5.1 limitation). Therefore, although multiple LAN adapters will work with older LAT implementations, you might still need to upgrade to Version 7.0 or higher of the OpenVMS operating system to obtain the LAT Version 5.3 protocol, which will correct this type of problem. (Note that this type of problem affects only those connections that are idle. An example of where this situation could arise is in an office environment if all users were to leave their systems at the same time, either at lunchtime or at the end of the workday.)

Large Buffers in Ethernet/FDDI Configurations  

The OpenVMS LAT software will attempt to use large buffers over any virtual circuit that comes in over an FDDI controller. This feature can cause problems if an alternate virtual circuit path must go through an Ethernet. LAT FDDI Ring and Large Buffers is an example of the configuration that can cause problems. 

Figure 6  LAT FDDI Ring and Large Buffers  
FDDI ring

In this diagram, it is possible for the two OpenVMS systems to communicate using large packets through the path described by controllers B and C. Large packets may exceed 1500 bytes of data (the maximum Ethernet message can contain 1500 bytes of data). If the path described by controllers B and C were to fail, it will not be possible for communication to continue through the path described by A and D.

The path described by controllers A and D pass through an Ethernet LAN segment. The messages that are routed through the 10/100 bridges cannot be larger than the maximum Ethernet message. Problems can occur because the OpenVMS LAT software cannot always detect this kind of configuration.

There are two ways to prevent problems with the previously described configuration. The first and easiest option is to create a logical LAT link using an Ethernet adapter (if either system has an Ethernet LAN adapter). This will force the message size negotiation to be no larger than the maximum sized Ethernet message.

If neither system has an Ethernet controller (thus making the first option not possible), the second option is to override the use of large buffer support (which is enabled by default) by using the new LATCP command qualifier, /[NO]LARGE_BUFFER. For example:

$ MCR LATCP SET NODE/NOLARGE_BUFFER
HP recommends that you use the SET NODE/NOLARGE_BUFFER command after all logical LAT links have been created and before the LAT node has been turned on. For example, note the order of the commands in LAT$SYSTARTUP.COM:
	$!
	$! Create each logical LAT link with a unique name and
	$! unique LAN address (forced with /NODECNET).
	$!
	$ LCP CREATE LINK FDDI_1 /DEVICE=FCA0 /NODECNET
	$ LCP CREATE LINK FDDI_2 /DEVICE=FCB0 /NODECNET
	$!
	$! Don't use large buffer support (force packet
	$! sizes to be no larger than what Ethernet can
	$! support).
	$!
	$ LCP SET NODE /NOLARGE_BUFFER
	$!
	$! Turn on the LAT protocol.
	$!
	$ LCP SET NODE /STATE=ON

go to previous page: Understanding the LAT Network Understanding the LAT Network
go to next page: Understanding the LATCP UtilityUnderstanding the LATCP Utility