WAN Subsystem Configuration and Management Manual

Overview of the WAN Subsystem
WAN Subsystem Configuration and Management Manual522463-011
2-14
Fault Tolerance and Fault Containment
Fault Tolerance and Fault Containment
Host-system connection fault tolerance is provided primarily by having two Ethernet
connections to the host for each WAN port and allowing the SWAN concentrator to use
either connection. Only one connection is needed for full functionality.
Fault-containment is provided by the independence of the three CLIPs, each of which
controls two WAN lines. If one CLIP fails, the other four WAN lines remain functional.
Additional fault-containment features include parity on the microprocessor DRAM and
the Ethernet anti-babble circuit that silences a babbling CLIP so that it cannot cause
the other two CLIPs to lose their Ethernet connections. The only component shared
across both of the Ethernet ports and the three CLIPs are the power supply and the
fan.
Path Switching
The preferred path might not always be available because of software or hardware
failures or because of the nature of the software and hardware configuration. However,
since the WAN subsystem and TCP/IP subsystems are designed to be fault-tolerant,
and since fault-containment is built into the SWAN concentrator, IOPs rarely encounter
a “NO PATH AVAILABLE” error. If this error is encountered, the SWAN concentrator
CLIP is unusable.
Table 2-8 describes the algorithm the ConMgr process uses to decide which path an
IOP will use to access the SWAN concentrator. When the conditions described in the
first column are met, the ConMgr process determines which path to use and how the
path will be accessed, as described in the second column.
Note. This algorithm is applicable for the TCP/IP subsystem only. The algorithm for Parallel
Library TCP/IP is shown in Table 2-9 on page 2-16.
Table 2-8. ConMgr Process PATH Switching Algorithm for TCP/IP (page 1 of 2)
When these conditions are met... The ConMgr process will...
A. If the ADAPTER preferred path
TCP/IP process
is started, and
has not failed, and
is primaried in the same proces-
sor as the primary IOP
Use the ADAPTER preferred path TCP/IP
process.
Select the IP address for the CLIP by using
the PATH A IP address.