To achieve smooth and stable execution of large-scale experiments and complex
analysis tasks, fault-tolerance is a top priority for a
simulation environment. The effect of an error, e.g., a failing system
call or a numerical error, has to be limited to the affected
experiment, with all other tasks in progress remaining
undisturbed
.
For the integration of external tools, the main concern lies on the detection of errors. To this end, several standard procedures are available in the VISTA/SFC environment to check the return code of system calls, to verify the existence of output files, and to examine the data generated. For instance, generated PIF files are automatically checked for containing the desired dopant distributions or geometry data. These procedures are encapsulated in a class of system tool objects automatically employed on every system call issued by the framework. If necessary, additional error handling routines can be defined for each system executable as well as for each application.
Robust operation is a an indispensable prerequisite for taking
full advantage of CAD support for technology development and analysis.
The reluctance usually encountered when proposing to perform
several hundred (or thousand) process flow simulation runs, each
involving a number of process steps and leading to the invocation of a
variety of simulation tools, is very much based on the fact that in
most cases, human interaction is required at many stages of the
simulation, making large experiments cumbersome and
time-consuming.
Therefore, response-surface modeling plays a very important role
in technology design and optimization, as it allows to reduce the
numbers of actually simulated design space points
[AAY
88] [MOA91] [CBK93] [KKW93]
[BM94] [GW94] [STA95] [SLW96];
in Chapter 8, the integration of RSM techniques is
treated in more detail.