Citation Context

...nd deadlines of dependent tasks [18]. Moreover, fault-tolerant scheduling algorithms can be classified into two groups: preemptive scheduling [27], [28] and nonpreemptive scheduling [25], [26], [32], =-=[33]-=-. In a preemptive scheduling approach, an executing task can be preempted by another task. In a nonpreemptive scheduling mechanism, running tasks cannot be interrupted during their executions and a ta...

Citation Context

...21], [22]. There are three variants of the primary–backup scheme: 1) the active-backup copy [23], [24]; 2) passive-backup copy [17]–[19], [21], [22]; and 3) primary–backup copy overlapping techniques =-=[25]-=-, [26]. In the active-backup copy schemes, if a primary copy (or a backup copy) finishes before its corresponding backup copy (or primary copy), the backup copy (or the primary copy) will not be termi...

Citation Context

...unaware of the architecture of the system. Therefore, two computer-sets must synchronize when they are both in well state, so as to realize the smooth switch if host computer fails. 4 Dong Liu, Chunyuan Zhang, Rui Li In single-task dual redundant systems, such as dual SCMs system, the synchronization can be achieved easily, because two computers’ input and output is simultaneous. However, it is different to our system, which runs multi-tasks operating system, since its data input and output is asynchronous. Moreover, with smaller data granularity, it gets more cost and worse extension ability [6, 7, 8]. In this paper, we present the task-level synchronization, which uses coarsely granular, different to signal-level synchronization and instruction-level synchronization. For system users, task is a command sent to the system by outer devices. And it determines the anticipant operation and behavior. For the system itself, task is data with specific format, saved in the task buffer. Each task possesses its own unique signification; and there is no explicit affiliation between two different tasks. Task number or time stamp marks a task. Furthermore, task is provided with a priority, which means ...

Citation Context

...n efficient fault tolerant technique in parallel and distributed systems. The three variants of this scheme include active backup copy [8], passive backup copy [9] and primary backup copy overlapping =-=[10]-=-. Generally speaking, backup copy is always preferred to be executed as passive backup copy because it can take the advantages of backup copy overloading and backup copy de-allocation technique to imp...

Citation Context

... of DRE systems. Prior research has also focused on passively replicated real-time fault-tolerant task allocation algorithms that deal with dynamic scheduling which exhibit extra overheads at runtime =-=[2, 53, 91, 153, 171]-=-. Prior research on static scheduling approaches has also focused on passively replicated real-time fault-tolerant task allocation algorithms that deal with only one processor failure [15, 114, 125]. ...