Title:
障害予測装置および障害予測プログラム
Document Type and Number:
Japanese Patent JP7424194
Kind Code:
B2
Abstract:
To provide a management device which uses machine learning, for predicting failures of a transmission path among each CPU in a parallel processing device.SOLUTION: A management device 3 includes a control unit 30, a storage unit 40, and an output unit 50. The control unit includes a prediction section 31 and an inventory management section 32. The storage unit includes a learning model 23, a prediction result DB41, an AOC inventory DB42, and a CPU inventory DB43. The learning model, in communication among a plurality of nodes included in a parallel processing device, is a machine-trained model using teacher data including adjustment parameters in communication waves representing a part of log information for the communication among each node and that is received in the nodes, and failure information for correct answers corresponding to the adjustment parameters. The prediction section sequentially inputs the adjustment parameters representing the part of plural pieces of the log information to be predicted, into the learning model, and outputs each failure prediction information from the learning model. The prediction section stores each failure prediction information corresponding to plural pieces of device state information, into a prediction result DB.SELECTED DRAWING: Figure 9
More Like This:
Inventors:
Junji Miki
Application Number:
JP2020079849A
Publication Date:
January 30, 2024
Filing Date:
April 28, 2020
Export Citation:
Assignee:
富士通株式会社
International Classes:
G06F11/07; H04L43/00
Domestic Patent References:
| JP7013983A | ||||
| JP2016038850A | ||||
| JP2006033249A |
Attorney, Agent or Firm:
Sakai International Patent Office