NonStop Networking Overview

Fault-Tolerance and Scalability on the NonStop System
When you have configured for scalability, the failure of one or more processors does not impact
the availability in any way:
• Availability of an application is not impacted by a failed processor because the application
process is replicated.
• Availability of Internet accessibility is not impacted by a failed processor because you have
running IP stacks configured on the running processors with the IP address still available so
clients do not have to talk to an alternate IP address.
• Recovery of scalability is not affected by a failed processor because when the processors are
reloaded and applications are restarted, the applications automatically resume their scaled
configuration without disruption to the single IP view. The re-instantiated applications start
using the IP address that had been configured previously.
Failover: Fault-Tolerant Connectivity
Failover allows a second interface to take over connections from an interface in a pair configured
for failover if the other interface fails.
Unlike other failover implementations, NonStop TCP/IPv6 and CIP failover do not require one of
the interfaces to act as a “hot standby” in anticipation of a failure. Both interfaces are active and
allow inbound and outbound network traffic to be distributed between them. Hence, in addition
to having fault-tolerance at the interface level by using failover, you also gain scalability when all
your interfaces are performing correctly.
For more information about failover, see the TCP/IPv6 Configuration and Management Manual
or the Cluster I/O Protocols (CIP) Configuration and Management Manual..
The CIP subsystem can be configured to provide failover between interfaces on a CLIM and can
provide failover between two CLIMs. Failover between two CLIMs results in a loss of TCP and SCTP
connections so applications need to reestablish those connections. However once sessions have
been reestablished, all other configuration aspects of the failed-over interface, are restored.
Logical Network Partitioning (LNP) and Providers
In the NonStop TCP/IPv6 environment, all applications have access to all interfaces (IP addresses)
unless you configure logical-network partitioning (LNP). The CIP environment shares this feature,
except you use Providers to restrict applications.
One of the most important differences between conventional TCP/IP and NonStop TCP/IPv6 and
CIP is that NonStop TCP/IPv6 and CIP have one manager process ($ZZTCP or $ZZCIP) and all
interfaces are associated with that single process. This difference influences how you configure
the subsystem. In conventional TCP/IP, you can have multiple TCP/IP processes, each having one
or more interfaces uniquely associated with them, as shown in Figure 8.
What’s Unique About Networking on Integrity NonStop Systems 19