ADAPTIVE JOB SCHEDULING USING NEURAL NETWORK PRIORITY FUNCTION - INTERNATIONAL BUSINESS MASCHINES CORPORATION

Title:

ADAPTIVE JOB SCHEDULING USING NEURAL NETWORK PRIORITY FUNCTION

Document Type and Number:

Japanese Patent JP3248657

Kind Code:

B2

Abstract:

PURPOSE: To allow a job scheduler to make a decision concerning the order and the frequency of access to a resource according to a substantially optimum delay cost function by utilizing a neural network as the delay cost function of the job scheduler.
CONSTITUTION: An operating system in a storage device 102 includes the job scheduler 122 and a performance monitor 123. The job scheduler 122 decides the order of serving a user job by CPU 101. The performance monitor 123 collects statistics showing the average of time necessary for a computer system 100 for serving a job in a different class. The statistics collected by the performance monitor 123 are stored in a performance data file 126. A neural network utility 124 and a neural network related data structure 125 constitute, learn and execute the neural network for establishing a delay cost function.

Inventors:

Joseph Philip Bigus

Application Number:

JP21709094A

Publication Date:

January 21, 2002

Filing Date:

September 12, 1994

Export Citation:

Click for automatic bibliography generation Help

Assignee:

INTERNATIONAL BUSINESS MASCHINES CORPORATION

International Classes:

G06G7/60; G06F9/46; G06F9/48; G06F9/50; (IPC1-7): G06F9/46; G06G7/60

Domestic Patent References:

JP6024647A
JP476660A

Other References:

【文献】IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS,Vol.E76-D,no.8(August 1993)P.947~955
【文献】Cihan Dagli,et.al,”A Neural Network Architecture for Faster Dynamic Scheduling in Manufacturing Systems”,Proceedings of the 1991 IEEE International Conference on Robofics and Automation,PP.2408~2413(1991)

Attorney, Agent or Firm:

Hiroshi Sakaguchi (1 person outside)

Previous Patent: SHUTTER DEVICE

Next Patent: ATTENDANCE CONTROL SYSTEM FOR PRIVATE BRANCH AUTOMATIC ELECTRONIC EXCHANGE