|
|
||
|
|||
|
|||
|
|||
|
|||
|
Why is MATT Being Developed? Testing of real-time systems presents many unique and challenging problems. Testing on target hardware requires solving problems including: the acquisition and installation of expensive target hardware, safety hazards, and the risk of costly hardware damage. Many times, target hardware cannot be taken off-line, or is simply unavailable for testing when needed. In the case of aeronautics and aerospace applications, the target hardware may not even exist when software testing is scheduled. Potential software problems can be too dangerous to test on target hardware until some level of testing is done in advance to assure safety. Even if safety requirements are met, a software defect could damage expensive equipment, such as a wind tunnel drive shaft.Developers often utilize simulation, prior to, or in conjunction with, hardware-software integration. Simulation presents another set of problems including generation and use of potentially huge data sets, time-consuming and error-prone output analysis, and testing time constraints. If a real-time system samples just 500 input values at 100-millisecond rate, then a simulation must supply 5,000 input values per second. A modest one-hour simulation of this system would require 18,000,000 values. If these values are floating point numbers, a 72-megabyte file is required for the input data. In addition, if the system produces 500 output values at the 100-millisecond rate, another 72 megabytes of output data would be produced. Analysis such an output file is simply overwhelming without the support of analysis tools. Further, many real-time systems include the requirement to execute for months or years without interruption. Testing for this length of time is not feasible within most project schedules. Even if a large amount of time is available, quantitative evaluation of domain testing faces some especially difficult problems. Input domains contain an overwhelming number of possible input values. For example, an input domain of 0.0 - 999.999 with an accuracy of just 0.001 contains 1,000,000 possible values. If a software system had just three input variables with this same domain, then 1,000,0003 combinations of input values are possible. Automated testing tools and measurement-based reliability evaluation methodologies are needed to improve software quality, increase testing productivity, and enhance management insight into process and product risk. These needs must be met without investment in expensive hardware and achieved across multiple projects.
|
|||
|
|||
| If you have any questions
or comments, we can be reached at the following addresses: Project Leader: Joel Henry MATT Customer Support Web Support: Spencer Wolny |
|
|
|
|
|||
