User guide

Improving Performance
Using ENFORM Efficiently
5–10 058058 Tandem Computers Incorporated
Specify Where ENFORM Builds Temporary Work Files
During an ENFORM session the query processor and the SORT process build
temporary work files while processing a FIND or LIST statement. Use generic files to
specify where these temporary files are built. Directing the query processor temporary
work files to a scratch disc is particularly useful. For example:
:ASSIGN QUERY-WORK-AREA, $SLOW
Place temporary files built by the SORT process on less used volumes. For example:
?ASSIGN QUERY-SORT-AREA, $QUIET
Spread Input/Output Demands Among Discs
Distribute data among different disc volumes to reduce the demands on any one disc
process. Distribute data by:
Partitioning files, especially those that are large and heavily used.
Moving data files to less used volumes.
Placing alternate key files on a different volume than the associated data file.
The improvement shown depends upon the organization of your system.
Alter Cache Size
The value of cache size is very application dependent. If main memory space is not
needed for other purposes, it can be valuable to make cache large enough to hold
every top level index block of each heavily accessed file, with a little space left over. In
practice, a reasonable starting allotment for cache is 30 pages per CPU containing disc
processes or 10 pages per disc process on a Tandem Nonstop II system.
Remember that cache uses main memory. The memory used for cache is not available
for other processes. When too much memory is used for cache, page faults result. As
cache size increases, the time needed to manage cache also increases. In a system
where files compete for cache (even very large cache), increasing cache size might not
improve performance.
Control the Size of the Target File
Control the primary and secondary extent sizes for the target file. If extents are too
large, disc space is wasted and the query might fail because an extent cannot be
allocated. If extents are too small, the target file might overflow the allocated space.
When the target file overflows, the query processor creates a new larger target file and
copies the old target file into the new one. This process of creating a new target file
and copying the old one degrades query performance, especially if the query processor
must do it more than once.