CWEA: Automated log retrieval for performance analysis of service-oriented scientific workflows

Fuente: Zenodo
Saved in:
Bibliographic Details
Main Author: Khaldi Ahanach, Elias el
Format: Recurso digital
Published: Zenodo 2019
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866901777138319360
author Khaldi Ahanach, Elias el
author_facet Khaldi Ahanach, Elias el
contents <p>A complex scientic workow often consists of many software tools or services, and those</p> <p>components are deployed on distributed infrastructures, e.g., when processing data from dis-</p> <p>tributed sources. The runtime behavior of the workow, e.g., monitored by the underlying</p> <p>infrastructure, is important for analyzing the provenance of the workow, in particular when</p> <p>the workow has unexpected performance issues or failure. However, it is very challenging</p> <p>to analyze the workow performance, due to the diculty in gathering and analyzing perfor-</p> <p>mance metrics across distributed infrastructures. Moreover, the workow provenance and the</p> <p>system logs are provided by workow management system and the underlying infrastructure,</p> <p>and they contain dierent information. In this thesis we aim to tackle this issue by propos-</p> <p>ing a tool that can automatically gather performance information (e.g., CPU, Memory and</p> <p>Network) for a given service-oriented workow. We assume the monitoring of the components</p> <p>and their machines is set up according to a common standard including often used tools like</p> <p>Docker, cAdvisor and the metric collector Prometheus. The proposed tool aims to integrate</p> <p>the dierent information, and provide methods for visual aid and evaluation of the system</p> <p>performance on dierent aspects. The prototype uses the provenance format, produced by</p> <p>Apache Taverna, which is a widely used workow management system in dierent elds of</p> <p>science. We validate our tool by simulating heavy resource load to create peaks in the per-</p> <p>formance data, and demonstrate how the tool detects and visualizes the irregularities in the</p> <p>experiment.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_3521576
institution Zenodo
language
publishDate 2019
publisher Zenodo
record_format zenodo
spellingShingle CWEA: Automated log retrieval for performance analysis of service-oriented scientific workflows
Khaldi Ahanach, Elias el
<p>A complex scientic workow often consists of many software tools or services, and those</p> <p>components are deployed on distributed infrastructures, e.g., when processing data from dis-</p> <p>tributed sources. The runtime behavior of the workow, e.g., monitored by the underlying</p> <p>infrastructure, is important for analyzing the provenance of the workow, in particular when</p> <p>the workow has unexpected performance issues or failure. However, it is very challenging</p> <p>to analyze the workow performance, due to the diculty in gathering and analyzing perfor-</p> <p>mance metrics across distributed infrastructures. Moreover, the workow provenance and the</p> <p>system logs are provided by workow management system and the underlying infrastructure,</p> <p>and they contain dierent information. In this thesis we aim to tackle this issue by propos-</p> <p>ing a tool that can automatically gather performance information (e.g., CPU, Memory and</p> <p>Network) for a given service-oriented workow. We assume the monitoring of the components</p> <p>and their machines is set up according to a common standard including often used tools like</p> <p>Docker, cAdvisor and the metric collector Prometheus. The proposed tool aims to integrate</p> <p>the dierent information, and provide methods for visual aid and evaluation of the system</p> <p>performance on dierent aspects. The prototype uses the provenance format, produced by</p> <p>Apache Taverna, which is a widely used workow management system in dierent elds of</p> <p>science. We validate our tool by simulating heavy resource load to create peaks in the per-</p> <p>formance data, and demonstrate how the tool detects and visualizes the irregularities in the</p> <p>experiment.</p>
title CWEA: Automated log retrieval for performance analysis of service-oriented scientific workflows
url https://doi.org/10.5281/zenodo.3521576