© Copyright IBM Corp. 2000 115
Chapter 7. Branch Extender
This chapter describes the functionality of Branch Extender and how it is
configured, and an example scenario illustrates the use of Branch Extender in
combination with HPR.
As the name implies, Branch Extender is designed for networks that can be
divided into distinct areas, such as separate branches of a large organization.
It works by separating out branch offices from the main APPN WAN backbone
network, for example, the organization’s headquarters network.
Each branch contains a node of the type called branch network node (BrNN),
which combines the functions of an APPN network node and an APPN end
To the backbone network, the BrNN appears as an end node, connected to its
network node server (NNS) in the backbone network. Because the BrNN
appears as an end node, it does not receive topology information from the
backbone network (topology information is transmitted only between network
The BrNN registers all resources in the branch with its NNS as though they
were located on the BrNN itself.This means that the nodes in the backbone
network can locate resources in the branch without having to be aware of the
separate nodes in the branch, thus reducing the amount of topology
information that must be stored. While a BrNN knows the topology of its local
branch, it knows nothing of the uplink topology.
To the branch network, the BrNN appears as a network node, acting as the
NNS for end nodes in the branch. Each node in the branch sees the rest of
the network as being connected through its NNS, in the same way as for a
standard NNS. When the BrNN can’t resolve a request for network services
locally, it refers the request to its NNS, which connects to other NNs using
normal CP-CP sessions. This can be thought of as default routing. That is
why a BrNN’s uplink to its NNS may be called the “default routing link”.
For directory services purposes, the BrNN appears to own all branch
resources. For route selection purposes, the BrNN appears to be the origin or
destination of all sessions involving branch LUs.
Chapter 7. Branch Extender 117
Figure 74. Sample “branched” network 2
Large enterprise networks can benefit greatly from advanced HPR functions
such as automatic resource discovery and route selection. However, the price
of dynamic network operation is the need to advertise the network topology or
search the network for resources. Prior to Branch Extender, any HPR node
that routed data between other systems (for example, CS/AIX connecting a
branch office to a host) was required to be a network node, and thus to send
and receive network control messages. In a small network, the overhead of
these network control messages is insignificant. But in a large network made
up of hundreds to thousands of relatively slow lines, even small overhead may
be unacceptable. See 7.2, “Topology databases” on page 118. IBM
introduced Branch Extender to enable such enterprises to gain the benefits of
HPR throughout their entire network without these concerns. It can support
end-to-end HPR connections all the way from computers in the branch to a
host system (this provides maximum flexibility in rerouting connections
around failed network components). The Branch Extender itself can also be
the endpoint of reroutable HPR connections and extend HPR reliability to