IBM BC-201 Network Router User Manual


 
Overview of IBM Networking
DLSw+
BC-209
Cisco IOS Bridging and IBM Networking Configuration Guide
78-11737-02
3x74 operates as if the acknowledgments it receives are from the 37x5. Router B looks like the 3x74 to
37x5. Because the frames do not have to travel the WAN backbone networks to be acknowledged, but
are locally acknowledged by routers, the end machines do not time out, resulting in no loss of sessions.
Enabling local acknowledgment for LLC2 has the following advantages:
Local acknowledgment for LLC2 solves the T1 timer problem without having to change any
configuration on the end nodes. The end nodes are unaware that the sessions are locally
acknowledged. In networks consisting of hundreds or even thousands of machines, this is a definite
advantage. All the frames acknowledged by the Cisco IOS software appear to the end hosts to be
coming from the remote IBM machine. In fact, by looking at a trace from a protocol analyzer, one
cannot say whether a frame was acknowledged by the local router or by a remote IBM machine. The
MAC addresses and the RIFs generated by the Cisco IOS software are identical to those generated
by the remote IBM machine. The only way to find out whether a session is locally acknowledged is
to use either a show local-ack command or a show source-bridge command on the router.
All the supervisory (RR, RNR, REJ) frames that are locally acknowledged go no farther than the
router. Without local acknowledgment for LLC2, every frame traverses the backbone. With local
acknowledgment, only data (I-frames) traverse the backbone, resulting in less traffic on the
backbone network. For installations in which customers pay for the amount of traffic passing
through the backbone, this could be a definite cost-saving measure. A simple protocol exists
between the two peers to bring up or down a TCP session.
Notes on Using LLC2 Local Acknowledgment
LLC2 local acknowledgment is enabled with TCP and DLSw+ Lite remote peers.
If the LLC2 session between the local host and the router terminates in either router, the other will be
informed to terminate its connection to its local host.
If the TCP queue length of the connection between the two routers reaches the high-water mark, the
routers sends Receiver-Not-Ready (RNR) messages to the local hosts until the queue limit is reduced to
below this limit. It is possible, however, to prevent the RNR messages from being sent by using the dlsw
llc2 nornr command.
The configuration of the LLC2 parameters for the local Token Ring interfaces can affect overall
performance. Refer to the chapter “Configuring LLC2 and SDLC Parameters” in this manual for more
details about fine-tuning your network through the LLC2 parameters.
The routers at each end of the LLC2 session execute the full LLC2 protocol, which could result in some
overhead. The decision to use local acknowledgment for LLC2 should be based on the speed of the
backbone network in relation to the Token Ring speed. For LAN segments separated by slow-speed serial
links (for example, 56 kbps), the T1 timer problem could occur more frequently. In such cases, it might
be wise to turn on local acknowledgment for LLC2. For LAN segments separated by a T1, backbone
delays will be minimal; in such cases, FST or direct should be considered. Speed mismatch between the
LAN segments and the backbone network is one criterion to help you decide to use local
acknowledgment for LLC2.
There are some situations (such as the receiving host failing between the time the sending host sends
data and the time the receiving host receives it), in which the sending host would determine, at the LLC2
layer, that data was received when it actually was not. This error occurs because the router acknowledges
that it received data from the sending host before it determines that the receiving host can actually
receive the data. But because both NetBIOS and SNA have error recovery in situations where an end
device goes down, these higher-level protocols will resend any missing or lost data. Because these
transaction request/confirmation protocols exist above LLC2, they are not affected by tight timers, as is
LLC2. They also are transparent to local acknowledgment.