TEYE is a HPC application feature analysis software, to extract the resource utilization information of large scale HPC cluster system, and feedback application feature in real time. TEYE exploits the potential computing capability of the application for bottleneck diagnosis, algorithm and parallel efficiency improvement. System resource optimization and computing performance improvement are top priority for TEYE.
System Parameters monitoring – Analysis & Diagnosis – System Optimization
Limited system resources are occupied by TEYE installation and usage. There are more than 40 Microarchitecture parameters (CPU, MEM, network and, file system), which could be accurately monitored and extracted by TEYE. These large amounts of raw data are provided for post analysis, diagnosis and optimization.
Based on knowledge and experiences on HPC application, Inspur TEYE offers more readable and flexible data graphics from raw data extraction.
System-level: analysing the running status of each node, to find out the key parameters.
Node-level: analysing the running status of parameters, to find out the bottleneck and output the analysis chart report.
Parameters-level: comparing with the same parameters from different nodes, analysing the load balance of nodes, and eliminates the bottleneck of nodes.
Distribution of data:
Displaying featured data and probability distribution of interval, so as to understand the feature of the application and the optimization direction.
Hot spot analysis:
Automatic analysis and identification of the points where performance data changes violently, analysing the feature of application at this point, providing the optimization.
Radar map of features:
According to the requirements of main indicators, generated radar map of features, finding the key parameters and performance bottlenecks.
Inspur application tuning Engineer combines graphic data with industry experience, to configure the most suitable solution for end user requirements, minimize the gap between hardware platform and application software, and maximize cluster performance and resource utilization.
In order to improve the parallel efficiency of application, optimize cluster performance, and minimize power consumption. TEYE outputs various graphic charts with comprehensive cluster Microarchitecture parameters, for application analysis and diagnosis, and further system configuration.
TEYE has been become an efficient tool of analysis and diagnosis for HPC administrators, application developer and system optimizer in a large scale HPC environment.