Fundamental to the design of SMART is the separation of importance information as expressed by user preferences from the urgency information as expressed by the time constraints of the applications. Prematurely collapsing urgency and importance information into a single priority value, as is the case with standard UNIX SVR4 real-time scheduling, results in a significant loss
of information and denies the scheduler the necessary knowledge to perform its job effectively. By providing both dimensions of information, the scheduler can do a better job of sequencing the resource requests in meeting the time constraints, while ensuring that even if not all time constraints can be met, the more important applications will at least meet their time constraints.
While SMART accounts for both application and user information in managing resources, it in no way imposes demands on either application developers or end users for information they cannot or choose not to provide. The design provides reasonable default behavior as well as incrementally better results for incrementally more information. By default, an end user can just run an application as he would today and obtain fair behavior. If he desires that more resources should be allocated to a given application, SMART provides simple controls that can be used to express that to the scheduler. Similarly, an application developer need not use any of SMART’s real-time programming constructs unless he desires such functionality. Alternatively, he might choose to use only time constraints, in which case he need not know about notifications or availabilities. When the functionality is not needed, the information need not be provided. When the real-time programming support is desired, as is often the case with multimedia applications, SMART has the ability to provide it.
