Specifications

40 Optimizing IBM Netfinity Servers for SAP R/3 and Windows NT
2.5.5 Redundant Network Interface Cards
A high-end server system such as the Netfinity 7000 with all its high-availability
features is useless if a single NIC fails. The server is running but not on the
network and the users cannot work with it. Implementing multiple NICs in your
SAP R/3 and database servers helps you to provide a higher level of availability
and eliminate the NIC as a potential single point of failure. All IBM Netfinity
models can operate with multiple network interface cards installed.
2.5.5.1 Duralink Failover
A particular solution to support an optimum availability for NICs is Adaptec’s
Duralink Failover integrated in their Fast Ethernet NICs. Duralink Failover is an
ingenious software solution that helps deliver high-availability network
connections.
The following NICs with this solution are supported in the Netfinity 7000:
1. Adaptec ANA-6911 single-port PCI Fast Ethernet adapter
2. Adaptec ANA-6944 four-port PCI Fast Ethernet adapter
Duralink Failover is a feature of the network device driver that monitors the
network adapter and the network infrastructure status. When enabled, Duralink
Failover monitors the network connection. In the event of a failure, it will move all
existing network traffic to a standby NIC or port.
Duralink Failover is available with both Adaptec’s single and four-port PCI Fast
Ethernet adapters. With these adapters, a pair of identical ports are configured in
the server. The two are designated as
primary and secondary . With single port
adapters, one is used as the primary port while another adapter must be installed
to provide the secondary port. If the server were to employ Adaptec’s four-port
adapter,
Quartet , for example, two ports may function as primary and the other
two as secondary.
Regardless of the configuration, when a failure is detected on the primary port,
Duralink Failover immediately disables the failed network port. Instantaneously,
the secondary port is brought online to keep the server
up . We call this sequence
of events
failover .
When failover occurs, SNMP traps are sent to network management stations and
error logs are updated through the Windows NT event logger. If the primary
adapter or port fails upon system boot, the system error log is updated and the
secondary adapter or port is enabled.
When enabled, failover will take place as a result of any of the following server
conditions.
• Link loss
• Watchdog timer expiration (a
keep-alive mechanism the adapter uses to poll
itself for proper operation)
• Abnormal hardware condition interrupt
• Abnormal send or receive counters, such as too many collisions or errors on
the segment