Distributed scheduling of tasks with deadlines and resource requirements

Abstract

In the design of distributed computer systems, the scheduling problem is considered to be an important one and has been addressed by many researchers. However, most approaches have not dealt with tasks' timing and resource requirements. In this paper, we describe a set of heuristic algorithms to schedule tasks that have deadlines and resource requirements, in a distributed system. These algorithms are dynamic and function in a decentralized manner. When a task arrives at a node, the local scheduler at that node attempts to guarantee that the task will complete execution before its deadline, on that node. If the attempt fails, the scheduling components on individual nodes cooperate to determine which other node in the system has sufficient resource surplus to guarantee the task. In this paper, four algorithms for cooperation are evaluated. They differ in the way a node treats a task that cannot be guaranteed locally. The random scheduling algorithm: The task is sent to a randomly selected node. The focused addressing algorithm: The task is sent to a node that is estimated to have sufficient surplus to complete the task before its deadline. The bidding algorithm: The task is sent to a node based on the bids received for the task from nodes in the system. Theflexible algorithm: The task is sent to a node based on a technique that combines bidding and focused addressing. Simulation studies were performed to compare the performance of these algorithms relative to each other as well as with respect to two baselines. The first baseline is the noncooperative algorithm where a task that cannot be guaranteed locally is not sent to any other node. The second is an (ideal) algorithm that behaves exactly like the bidding algorithm but incurs no communication overheads. The simulation studies examine how communication delay, task laxity, load differences on the nodes, and task computation times affect the performance of the algorithms. The results show that distributed scheduling is effective even in a hard real-time environment and that the relative performance of these algorithms is a function of the system state.

Dynamically mapping tasks with priorities and multiple deadlines in a heterogeneous environment

“…The concept of moving a task if its deadline may not be satisfied (presented in [30]) is used in one of the heuristics in our research. However, the environment in our research is different because it includes task with priorities and multiple deadlines, heterogeneous machines, and a more complex performance metric, all of which complicate the scheduling problem.…”

“…The environment in [30] has randomly arriving tasks with a hard deadline. The concept of moving a task if its deadline may not be satisfied (presented in [30]) is used in one of the heuristics in our research.…”

Dynamic mapping in a heterogeneous environment with tasks having priorities and multiple deadlines

“…The environment in [20] is similar to the environment considered in our research in that [20] has randomly arriving tasks with a hard deadline. The idea in [20] to move a task if its deadline may not be satisfied is used in one of the heuristics in our research.…”

“…The idea in [20] to move a task if its deadline may not be satisfied is used in one of the heuristics in our research. However, the environment in our research includes task with priorities and multiple deadlines, and heterogeneous machines, all of which complicate the scheduling problem.…”

Probabilistic job scheduling for distributed real-time applications

“…Most tasks in such systems have strict execution deadlines. Depending on the strictness of the execution deadline, tasks are categorized as critical and essential 15], 16]. If missing a task's deadline is catastrophic, then the task's deadline is considered to be hard, and the task is categorized as critical.…”