Availability Guide for Application Design

Overview of Server and Network Fault Tolerance
Availability Guide for Application Design—525637-004
2-4
Parallel Hardware Components
Parallel Hardware Components
HP’s parallel architecture provides fault tolerance while remaining cost-effective. Unlike
other fault-tolerant schemes, the HP approach is not based on lock-stepped
components or idle “hot” standby components that are used only for the duration of a
failure. The HP approach makes use of all available modules, thereby maximizing cost-
effectiveness. Even if a component fails, the efficiency of the server is still at least as
high as a system built of the same components around lockstep design or hot standby
components.
HP’s parallel architecture is more cost-efficient the more parallel modules you have.
For example, a 4-processor server running at 64-percent utilization becomes a
3-processor server running at 80-percent utilization if a processor failure occurs. A
16-processor server, however, can safely run at 75-percent utilization because, even
following failure of a processor, a resulting 15-processor server still runs at 80-percent
utilization.
The latest NonStop server is the HP Integrity NonStop NS-series server (also known
as a TNS/E system) which uses the same ServerNet fabric and hardware for I/O
communications as the S-series, but has a different method of checking the integrity of
the individual processors. Because the NS-series builds on and extends the strengths
of the two previous architectures it will be covered at the end of this section, in The HP
Integrity NonStop NS-Series Server Provides Another Level of Availability on
page 2-23.
Figure 2-2 shows the parallel architecture of part of a typical HP NonStop S-series
server. Figure 2-3 on page 2-6 shows the parallel architecture of part of a typical HP
NonStop K-series server. The S-series server uses ServerNet fabrics in place of both
the bus and input/output channel portions of the older K-series server architecture.
The following paragraphs explain how the parallel architecture components of each
server provide fault tolerance.