Fault prediction method for fault log of high-performance computing system

Abstract

The invention discloses a fault prediction method for a fault log of a high-performance computing system, and the method comprises the following steps: obtaining the fault log data of the high-performance computing system, and obtaining a fault time sequence through the analysis of the fault log data, and the fault time sequence is suitable for an LSTM model; adopting a K-means algorithm to carryout clustering processing on the fault types contained in the fault log data; building an FD-LSTM model based on the fault time sequence; and based on the FD-LSTM model, predicting a fault occurrencenode position and fault advance time of a clustering result of each fault type, and carrying out statistical analysis on a prediction result according to a system architecture. According to the invention, through fault classification prediction, the fault analysis accuracy of a high-performance computing system can be effectively improved, the operation and maintenance efficiency is enhanced, andthe reliability of the system is effectively improved.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a fault prediction method for high-performance computing system fault logs. Background technique [0002] In order to pursue higher simulation accuracy and obtain more calculation details, scientists increasingly rely on high-performance computers to process unprecedentedly large data sets and complex simulations. High-performance computers have developed rapidly from the initial single-chip system to the cluster system (Cluster) with tens of thousands of processors; and until now, the main means of improving computer performance is still to increase the number of processors. Lead to the rapid expansion of the scale of high-performance computers. At the same time, there are higher requirements on the ability of the system, including software and hardware, to deal with sudden error events, that is, the fault tolerance capability. In particular, the increase in processing...

AI Technical Summary

Problems solved by technology

While some failures are detectable and obvious, such as kernel panics, most anomalies are not easy to track down

Which component will fail, and how it will affect the system, is unclear

Abnormal symptoms observed in the system may or may not reflect the exact root cause, for example, a kernel panic may be caused by a Luster file system error or a hardware machine check exception

Images