TCP/IP (Parallel Library) Configuration and Management Manual

Configuring Parallel Library TCP/IP for Complex and
Heavy-Use Environments
HP NonStop TCP/IP (Parallel Library) Configuration and Management Manual—522271-006
3-29
Parallel Library TCP/IP for Complex, Heavy-Use
WAN Environments
Parallel Library TCP/IP for Complex, Heavy-
Use WAN Environments
Parallel Library TCP/IP provides improved scalability for the SWAN subsystem.
Whereas with conventional TCP/IP, traffic for a given path had to go through the
processor which contained the conventional TCP/IP process, with Parallel Library
TCP/IP, you can configure the WAN subsystem so that not only the workload of a WAN
I/O process but also the workload of TCP/IP done on its behalf is spread across all the
processors in which you have configured the WAN I/O processes.
In conventional TCP/IP, there was a one-to-one correspondence between the TCP/IP
process and a SLSA LIF/PIF. Traffic destined for that TCP/IP process would all flow
through a single processor (potentially creating a bottle-neck). With Parallel Library
TCP/IP, there is a one-to-any correspondence between the TCP/IP stack and all the
LIFs/PIFs. So now you can place your WAN I/O processes in any processor and you
can associate them with any LIFs/PIFs. Traffic now goes directly through the
processors in which the WAN I/O processes have been configured, eliminating the
interprocessor hop formerly required to get to the TCP/IP process servicing the
LIF/PIF.
In addition to increased scalability, Parallel Library TCP/IP also offers a feature called
Ethernet failover. This provides another layer of fault-tolerance at the Ethernet adapter
level, from which the SWAN subsystem and its client-I/O processes can benefit.