Real-Time Virtual Resources

University Of Texas At Austin, Austin TX

Investigators

Abstract

Aloysius K. Mok CCR-0207853 "REAL-TIME VIRTUAL RESOURCES" As embedded systems become more complex, a typical embedded system will probably involve a mix of soft and hard real-time applications that share the same embedded run-time platform. In general, an application may consist of one or more tasks. For example, on a hand-held sensor-computer, a user may want to fix his own location with respect to a target landmark and also be warned of the power consumption status of the sensor-computer. While the power consumption warning task may have very loose timing constraints, the location fixing task has fairly stringent timing and other QOS (quality of service) requirements. Ideally, an application developer should be able to write his application program as if it were running on a dedicated computer and not have to worry about interference from other applications. This illusion can be maintained if global knowledge of the QOS requirements of all the applications is accessible to the operating system scheduler, so that a global schedulability analysis can be performed to ensure that every task meets its QOS requirements. However, this is not possible in the environment of "open systems" where the operating system may not know about the timing requirements of all the application tasks, as is the case with most commercial embedded and real-time kernels. A simplistic solution is to assign a higher priority to timing- critical applications and let the non-real-time applications run only when the timing-critical ones are finished. This approach is viable only for the simplest embedded systems where there is only one real-time task and there is no interaction among the real-time and the non-real-time tasks. If all the applications are to various degrees timing-critical (hard or soft), then the distinction between real-time and non-real-time applications becomes blurry. In the sensor-computer example, if the power consumption warning task is assigned a low priority because of its laxer timing constraint, then it is possible that power may run out while the high-priority location fixing task is monopolizing the CPU. This may in turn cause a sensor reset that may affect the validity of the location fixing result. The goal of this project is to enable the design of robust embedded real-time systems that must function in an open systems environment. The innovation we are investigating is an elegant software abstraction called RTVR (real-time virtual resource). This abstraction allows application programmers to design embedded real-time systems as if each application had exclusive access to a set of dedicated physical resources which provide service at a roughly constant rate. The key to the RTVR concept is that of a "delay bound" parameter on the rate of service provision. The delay bound specification enables a programmer to specify the jitter allowance of his/her application. As a result, we can simplify the verification of the timing correctness of individual applications as if there is no timing interference among application tasks. The RTVT concept also simplifies the issue of guaranteeing the timing/QOS requirements of a mixture of hard and soft real-time applications. To realize this abstraction, we must make major advances in both theory and engineering in operating system design and real-time scheduling theory. Of particular interest to this project are the issues of partition composition and the hierarchical decomposition of multiple real-time virtual resources.