BenchMark Guidelines

Previous Next

Generally speaking, USoft BenchMark can test the following:

· Startup time of the application


· Startup time of info windows


· Transaction time


· Batch-processing time


· Workflow time

How representative the test model is depends upon a few items, which influence the performance. These items are:

· Memory use, which depends upon:
· Number of controls appearing on the screen; the more (painted) controls, the harder it is for the system to function


· Number of records that will be shown.


· Buffer height.


· Size of the resident cache.
· Startup time of the application, which depends upon:
· Save Workspace Options used (windows on the screen, in memory, or cached)


· Number of windows in the initial picklist


· Connect time of the application to the database (with respect to modem connections)


· Resource and application files, either on the network or locally (application files include resource files, CON, EXT, ESI, bitmaps, etc.)
· Startup time of info windows, which depends upon:
· Type of resident cache


· Whether the info windows are in the cache or not


· Size of the screen


· Number of records


· Number of columns


· Number of embedded children


· Number of ESI definitions


· Startup actions

The following items also influence performance:

· Use of joined columns (whether it is optimized or not)


· Use of additional fields


· Use of extra queries


· Number of relationships ( when an info window is opened the relevant table and all the direct relationships are read in from the conceptual file)


· Use of subtypes ( this can also positively influence the performance)


· Number of restrictive and manipulation constraints


· Hardware and software configuration such as:
· Swap space tuning


· Amount of memory


· Network map (throughput)


· Graphics accelerator


· Disk access (32 bit)


· Network driver software + configuration


· Database tuning (extra indexes )


· Whether you are running from flat files or not

Related Topics:

Resident Cache

Transaction Performance

Ignoring Import Errors