Fast Scheduling of Distributable Real-Time Threads with Assured End-to-End Timeliness

Scheduling Dependent Distributable Real-Time Threads in Dynamic Networked Embedded Systems

“…A server is correct if it does not fail; it is faulty if it is not correct. DQBUA tolerates up to N − 1 client failures and up to f s max ≤ n/3 server failures (see [6]). The actual number of failures is denoted as f s ≤ f s max for servers and f ≤ f max where f max ≤ N − 1 for clients.…”

“…In [6], we develop QBUA, a scheduling algorithm for real-time threads in partially synchronous systems. Here, we extend QBUA to handle resource dependencies and precedence constraints, we call the resulting algorithm DQBUA.…”

“…This arbitration reduces thrashing by minimizing the number of instances of DQBUA that are started to handle the same or concurrent scheduling events. Due to space limitations, we do not reproduce the details of the quorum arbitration algorithm, see [6] for details.…”

“…In this paper, we consider the problem of scheduling dependent threads in the presence of the previously mentioned uncertainties. Past efforts on thread scheduling (e.g., see [6] and references therein) can be broadly categorized into two classes: independent node scheduling and collaborative scheduling. In the independent scheduling approach, threads are scheduled at nodes using propagated thread scheduling parameters and without any interaction with other nodes.…”