Unlock instant, AI-driven research and patent intelligence for your innovation.

A Fast Task Allocation Method to Avoid Deadlock on Heterogeneous Resource System

A technology for task allocation and deadlock avoidance, applied in the direction of resource allocation, multi-programming devices, etc., can solve the problem of communication dependencies without considering the priority of tasks, minimize the total overhead, reduce the scale, and improve the efficiency Effect

Active Publication Date: 2016-08-10
CHONGQING UNIV
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Some existing techniques are used to minimize task overhead, but they do not consider the dependencies of tasks, the communication dependencies between tasks, and the security of local scheduling to avoid deadlocks

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A Fast Task Allocation Method to Avoid Deadlock on Heterogeneous Resource System
  • A Fast Task Allocation Method to Avoid Deadlock on Heterogeneous Resource System
  • A Fast Task Allocation Method to Avoid Deadlock on Heterogeneous Resource System

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0016] The idea of ​​the present invention is to merge the nodes of the task data flow graph into node clusters. However, if the new task data flow graph composed of node clusters has loops, deadlock may occur during task scheduling. In a directed acyclic task data flow graph, the necessary condition for generating a loop after merging two nodes into a node cluster is that there is a path greater than or equal to 2 between the above two nodes. After performing geometric operations on the adjacency matrix, a new matrix can be obtained. Through the matrix elements in the new matrix, it can be determined whether there is a path of a certain length between nodes. In this regard, we gave a series of definitions, theories, and carried out mathematical proofs.

[0017] It is an existing technical means to use task data flow graph to represent an application task. References include: ShaoZL, Zhuge QF, Xue C, Sha HM. Research on efficient allocation and scheduling on heterogeneous dsp sy...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention relates to a quick task allocation method avoiding deadlock on a heterogeneous resource system. The quick task allocation method comprises the following steps of: (1) representing a specific application by using a task data flow diagram, and geometrically calculating by using an adjacent matrix of the task data flow diagram to acquire a new matrix, wherein the new matrix is used for representing whether the length of the longest path between nodes is greater than or equal to 2; (2) selecting node pairs with the longest path length between nodes smaller than 2 and greatest communication data volume according to the new matrix, merging the node pairs into a node cluster, updating the new matrix, repeating the node merging process, and reducing the scale of the task data flow diagram to an intended target; and (3) allocating the task data flow diagram to a heterogeneous resource set by using an integral linear programming equation set. The quick task allocation method has the advantages of being provided with task execution sequence dependency relationship and communication dependency relationship among tasks and has the characteristic of avoiding deadlock, and the total expense of the heterogeneous resource system is minimized.

Description

Technical field [0001] The invention relates to a large-scale task distribution on a heterogeneous machine cluster, and in particular to a fast task distribution method for avoiding deadlock on a heterogeneous resource system. Background technique [0002] Modern high-performance scientific computer systems and common commercial computer systems are almost all based on the computer system structure of heterogeneous clusters. In a heterogeneous computer system, processors of the same type are interconnected to form a cluster, and different clusters are connected through heterogeneous communication links. Heterogeneous clusters can provide very high computational efficiency at the cost of huge overhead. [0003] In a heterogeneous resource system, large-scale tasks are completed on various types of heterogeneous resources, and a large number of different types of heterogeneous resources must be coordinated to complete a large-scale task. figure 1 Shown is a schematic diagram of a he...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F9/50
Inventor 沙行勉诸葛晴凤孙群燕刘铎
Owner CHONGQING UNIV