Monitoring organization method for high-performance computing application characteristics

Abstract

The invention provides a monitoring organization method for high-performance computing application characteristics. The method comprises the following steps: defining a management node and a computing node as a monitoring control end and a monitoring client on the basis of a high-performance computing application according to the characteristics of a high-performance computing hardware platform architecture; completing management, data receiving, data review and storage work of the client by using the control end; receiving a command of the control end, completing acquisition work of running characteristic parameters within a specified time interval, and transmitting to the control end according to the command of the control end by using the client. According to the method, the characteristic parameters of five links including inputting, storing, processing, transmitting and outputting during running of the application are monitored and extracted comprehensively, so that the data acquisition quantity is large, and the resource demands of the application on the aspects of CPU (Central Processing Unit) occupation, internal memory capacity, internal memory throughput, input / output, network data exchange and the like as well as the usage levels of relevant computing resources are shown sufficiently. Moreover, the method further has the characteristics of low system resource occupation rate, large monitorable cluster scale, repeatable inquiry of data, and the like.

Description

technical field [0001] The invention relates to the technical field of computer applications, in particular to a method for monitoring and organizing application characteristics of high-performance computing. Background technique [0002] With the continuous progress of human society and the development of science and technology, people not only have a wider understanding of the natural world, but also have an increasingly urgent need to explore the unknown world. This has led to a sharp increase in the amount of information and data held by humans, and at the same time, these massive information and data need to be analyzed and processed in a timely manner. For example, a large astronomical radio telescope array can generate more than 100GB of cosmic microwave data in one second, and these data need to be analyzed in time; Data is also measured in terabytes; in addition, fields such as human genome engineering, oil exploration, weather forecasting, etc. have also put forwa...

AI Technical Summary

Problems solved by technology

[0005] Judging from the current situation, there is no free system monitoring tool in the industry that can monitor hardware system microarchitecture indicators in real time, such as: CPI, real-time memory bandwidth, system real-time floating-point calculations, etc.

And in other non-microarchitecture indicators, the support of existing monitoring tools is not very good, a typical example is the IB network (rdma)

In addition, in most cases, other monitoring tools cannot flexibly meet specific requirements, such as nmon cannot meet accurate multi-node monitoring; Ganglia is not very good at data collection intervals and storage

Images