Workshop Report: 1991 Workshop on Architectural Aspects of Real-Time Systems, San Antonio, Texas, U.S.A.
✍ Scribed by C. M. Krishna; Y. -H. Lee
- Publisher
- Springer US
- Year
- 1992
- Tongue
- English
- Weight
- 153 KB
- Volume
- 4
- Category
- Article
- ISSN
- 0922-6443
No coin nor oath required. For personal study only.
✦ Synopsis
A large number of papers dealt with the allocation and scheduling of tasks to meet hard deadlines. In their paper, "Multiple Resource Allocation and Scheduling for Multiprocessor Real-Time Distributed Systems," Mosse et al. present a strategy for task allocation and scheduling. The application is characterized by a computation graph, where the nodes represent tasks and the edges represent the dependencies between tasks. The resources are divided into active (e.g., processors) andpassive (e.g., buses) categories. With this underlying model, tasks can first be allocated to individual processors, and then scheduled on that processor.
In their paper, "The DRTREE Architecture for Distributed Hard Real-Time Systems," Audsley et al. show how to use simulated annealing to allocate tasks to processors. The feasibility of each allocation (i.e., whether deadlines can be met on each procesor) is determined by the standard deadline monotonic approach.
In "SpringNet: A Scalable Architecture for High Performance, Predictable and Distributed Real-Time Computing," Stankovic et al. discuss an architecture composed of small-scale multiprocessors connected via replicated memory. Connection to the memory is by means of fiber optics. The overall goal is to support a priori guarantees of timely execution for critical tasks, on-line guarantees for other non-critical tasks, and efficient resource allocation.
In "A Multicomputer for Real-Time Software Constructed from Reusable Components," Welch and Stoyenko outline their Architecture for Reusable Components (ARC), which is meant to efficiently support programs composed from reusable software components. The processors are RISC machines, with special-purpose on-chip circuitry for interprocessor communication.
In "Multiprocessor Support for Real-Time Fault-Tolerant Scheduling," Oh and Son discuss the use of multiple copies of tasks to ensure that real-time requirements are met despite the failure of a subset of the computer system.
In '~rchitectural Impact of FDDI Network on Scheduling Hard Real-Time Traffic," Agrawal et al. consider how to guarantee that messages will be delivered by a prespecified deadline in FDDI token-ring networks. The FDDI protocol does not support a global priority allocation. The authors suggest a simple scheduling technique that guarantees deadlines of critical messages.