FINJ: A Fault Injection Tool for HPC Systems

Netti, Alessio; Kızıltan, Zeynep; Babaoğlu, Özalp; Sîrbu, Alina; Bartolini, Andrea; Borghesi, Andrea

doi:10.1007/978-3-030-10549-5_62

Cited by 10 publications

(20 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Fault Programs. All the fault programs used to reproduce anomalous conditions on Antarex are available at the FINJ Github repository [15]. As in [17], each program can also operate in a low-intensity mode, thus doubling the number of possible fault conditions.…”

Section: Features Of the Datasetmentioning

confidence: 99%

See 1 more Smart Citation

Online Fault Classification in HPC Systems Through Machine Learning

Netti

Kızıltan

Babaoğlu

et al. 2019

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

As High-Performance Computing (HPC) systems strive towards the exascale goal, studies suggest that they will experience excessive failure rates. For this reason, detecting and classifying faults in HPC systems as they occur and initiating corrective actions before they can transform into failures will be essential for continued operation. In this paper, we propose a fault classification method for HPC systems based on machine learning that has been designed specifically to operate with live streamed data. We cast the problem and its solution within realistic operating constraints of online use. Our results show that almost perfect classification accuracy can be reached for different fault types with low computational overhead and minimal delay. We have based our study on a local dataset, which we make publicly available, that was acquired by injecting faults to an in-house experimental HPC system.

show abstract

Section: Features Of the Datasetmentioning

confidence: 99%

“…This number does not include the metrics collected by the procinterrupts plugin, which were found to be irrelevant after preliminary testing. All the scripts used to process the data are available on the FINJ Github repository [15].…”

Section: Creation Of Featuresmentioning

confidence: 99%

Online Fault Classification in HPC Systems Through Machine Learning

Netti

Kızıltan

Babaoğlu

et al. 2019

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

show abstract

“…-FINJ Fault injection is a tool for high performance computing [14] in Python that is implemented as an oriented object, with high level programming language, that is used on many operating systems majors.…”

Section: Swifi Toolsmentioning

confidence: 99%

“…To perform variations of HPC systems and observe their behavior, Netti et al have present FINJ and implemented in Python in a high-level fault injection tool for fault injection and monitoring in the systems [14]. FINJ has been designed implemented to an open source of easy use in Python tool at HPC systems.…”

Section: Analysis Of Literaturementioning

confidence: 99%

“…Based on a high-level fault injection tool, Netti et al [14] has designed and implemented a FINJ tool, an easy-to-use open source Python tool for HPC systems as a fault injection target. Authors in [15] and [16] have calculated an important tool for fault predictions that is the probability of specified system failure modes, such as stop behavior failure and painful failures of system safety.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Review of Dependability Assessment of Computing System with Software Fault-Injection Tools

Alazawi¹,

Al-Salam²

2019

Journal of Southwest Jiaotong University

View full text Add to dashboard Cite

The importance of systems dependability is the main motivation for developing fault injection methods (because fault injection is one of the most important ways to assess system dependability). Dependability assessment automatically generates mathematical models of reliability based on simple representations of the parameters of the functions of the system components, service and logistics functions, subsystem order structures, and critical functions. If necessary, it is also possible to develop a detailed mathematical model for various aspects of the stability system behavior. Package evaluation functions include accurate and approximate solution methods that can be used to quantify the reliability of components, subsystems and operations, maintenance functions, and availability with respect to potential project design. In this research, several fault injection tools and software system dependability assessment are rewired. The general method of dependability assessment for software systems under testing is a fault injection tool. Dependability assessment based on the fault injection tool can be done by inserting the faults into a system and monitoring that system for determining its behavior and response. Practically, experimented fault injection techniques can be grouped into fault injection implementation by hardware, software, simulation, emulation, and hybrid fault injection techniques. This study aims to measure the weakness that affects dependability attributes by presenting a survey on the implementation of software fault injection approaches for three major levels (interface, distributed systems, and operating systems levels) with dependability applications and an overview of the different injection tools.

show abstract