Availability Guide for Application Design

Overview of Server and Network Fault Tolerance
Availability Guide for Application Design—525637-004
2-17
Fault Tolerance in a Client/Server Network
•
Data structure inconsistency
•
Invalid input values (for example, invalid array indexes)
•
Noninitialized or null pointers
•
Procedure call error returns
TFDS isolates software problems and provides automatic failure data collection,
diagnosis, and recovery services.
A typical action of TFDS would be to stop the errant processor, automatically reload it,
and compare the reported error against a log of known error conditions. If a similar
error has previously occurred, TFDS simply updates counters in the log. If the error is
new, TFDS is able to report the error to an operator for further analysis.
Fault Tolerance in a Client/Server Network
A continuously available server does not guarantee that an application will always be
available to the end user if other hardware or software is involved. Specifically, in a
client/server application, you must consider the effects of failures in the local area
network (LAN) and in the client processor itself.
HP has enhanced its leadership in highly available systems by extending its fault-
tolerant, parallel architecture to the LAN:
•
On S-series or NS-series systems, with ServerNet Wide Area Network (SWAN)
concentrators for dedicated LAN support, Ethernet 4 ServerNet adapters (E4SA)
and Gigabit Ethernet ServerNet Adapters (GESA) for public LAN support
•
On K-series systems, with the NonStop Access for Networking product
These products provide a base for highly available client/server applications.
This subsection extends the discussion under Fault Tolerance in the Server System by
showing how you can improve the availability of client and network hardware. It
discusses the following topics:
•
Additional availability problems in client/server networks
•
Benefits of a continuously available server
•
A fault-tolerant LAN
You can increase the availability of a client/server application by taking specific
application design steps. Refer to Section 5, Increasing the Availability of Tuxedo
Applications, or Section 6, Availability in the Pathway Transaction-Processing
Environment.
Additional Availability Problems in Client/Server Networks
The growth in distributed applications in recent years has made fault-tolerant design
more complex. Fifteen years ago, the typical computer application involved dumb
terminals attached to a mainframe computer system. The concept of availability simply
meant keeping the mainframe running. Today, many applications are spread among
client and server systems from various vendors. For a continuously available