Toward a realistic task scheduling model

Abstract

Abstract-Task scheduling is an important aspect of parallel programming. Most of the heuristics for this NP-hard problem are based on a very simple system model of the target parallel system. Experiments revealed the inappropriateness of this classic model to obtain accurate and efficient schedules for real systems. In order to overcome this shortcoming, a new scheduling model was proposed that considers the contention for communication resources. Even though the accuracy and efficiency improved with the consideration of contention, the new contention model is still not good enough. The crucial aspect is the involvement of the processor in communication. This paper investigates the involvement of the processor in communication and its impact on task scheduling. A new system model is proposed based on the contention model that is aware of the processor involvement. The challenges for the scheduling techniques are analyzed and two scheduling algorithms are proposed. Experiments on real parallel systems show the significantly improved accuracy and efficiency of the new model and algorithms.Index Terms-Parallel processing, concurrent programming, scheduling and task partitioning, processor involvement, heterogeneous system model.

Scheduling in a dynamic heterogeneous distributed system using estimation error

“…Many scheduling algorithms (other than the most trivial) utilize knowledge of the available system resources and the tasks to be processed when deciding to allocate a task to a processor [3,8,10,28,33,36,37,42]. How to best generate this knowledge is an open problem [37], which is dealt with in different ways.…”

“…The accuracy of these estimations becomes particularly important in distributed systems, where users must pay for processing, where demand for computational resources outstrips supply, or where a problem must be processed as quickly as possible. The most common form of estimating task execution times is by benchmarking a task or set of tasks offline in advance [3,8,33,36,37]. The heterogeneous and non-dedicated nature of the resources in a loosely-coupled distributed system means that this type of estimation can introduce a large amount of error, between the actual execution time and the estimated execution time of a set of tasks.…”

“…Other algorithms require the user to define the length of time that a problem is expected to take [3,8,33,36,37]. Many require a directed acyclic graph with task and communication information, and precedence constraints given in advance [3,8,24,26,25,33,36,37].…”

“…Many require a directed acyclic graph with task and communication information, and precedence constraints given in advance [3,8,24,26,25,33,36,37]. Communication costs are also not properly considered by many algorithms, for example all communication links are assumed to be homogeneous [41], it is assumed communication and computation can take place simultaneously [8], or it is assumed that there is instantaneous message passing [43].…”

“…Sinnen et al [33] look at a processor's involvement in communication and show that considering this involvement, when scheduling, leads to more efficient resource utilization in real-world distributed systems. Cohen et al [5] focus on scheduling the communication between processors, to minimize the communication overhead in a distributed system.…”