Token Ring Adapter Mirroring

Introduction

In a modern corporate network, server connections are critical. However, like other parts of the network, they are also vulnerable. Server connections depend on secure cabling connections and this means that, as soon as a cable gets damaged or inadvertently pulled loose, the server connection can disappear.

Adapter mirroring is a way of protecting server connections. It involves fitting the server with a second adapter which - whenever necessary - can take over from the first one automatically and without disruption to network users.

Mirroring

In a server with mirrored adapters installed, one adapter is called the Master or Primary adapter and the other is called the Slave or Standby adapter. The Master adapter opens onto the network using a Locally Administered Address (LAA). If the adapter's driver detects a problem with its connection to the hub, it shuts down the Master adapter and opens the Slave adapter onto the ring. The Slave adapter opens using the same LAA as was previously used by the Master adapter.

Once the changeover has taken place, the network administrator can fix the problem that made it necessary. Then, if for any reason the connection provided by the Slave adapter fails, the Master will automatically take over again.

Connecting the Master and Slave adapters to the ring

For effective adapter mirroring, you should connect the adapters to the same CAU but by using different cabling paths. This way the cables for both adapters are less likely to damaged in a single incident. If you do use a different CAU for each adapter, make sure that the two CAUs are not separated by a source routing bridge. Otherwise client PCs will not be able to communicate with the server by using the same addressing information (despite the server's LAA being the same) as they used before the server performed its adapter changeover.

Events that result in an adapter changeover

If the device driver on a mirrored adapter detects a hardware problem, for example, a lobe cable fault or a signal loss, it automatically changes over to the alternative adapter. In the event of a lobe cable fault, an adapter always goes through a fault-recovery procedure. With adapter mirroring, however, you can set a timer that causes the device driver to change adapters before the fault-recovery cycle is complete. This minimises any pause in network services that users might otherwise experience. (Note that, under Windows NT, the changeover process is governed by the NT system, and not by the device driver.)

The device driver will also start using the alternative adapter if the ring is beaconing persistently. You can set a timer to govern this as well. After the period you specify, the driver will change adapters to enable the server to try communicating over a different cabling path.

Finally, you can configure the driver to change adapters if it becomes isolated on the ring. By default, Madge drivers do not change adapters in this case (it is normal for an adapter to be the only member of a ring at times, for example, when the server starts up). However, you can set a timer which causes the driver to change adapters after a certain period of isolation. This ensures that, for example, the server connection survives even if the CAU has detected a cabling problem and isolated the adapter the server is using.

Adapter Mirroring and Madge

The Madge driver MMIRROR.NLM supports adapter mirroring for Smart Ringnodes in NetWare servers.

The Madge NDIS3 Miniport driver supports adapter mirroring for Smart Ringnodes in Windows NT servers.