Managing the ATLAS Grid through Harvester
ATLAS Computing Management has identified the migration of all computing resources to Harvester, PanDA’s new workload submission engine, as a critical milestone for LHC Run 3 and 4. This contribution will focus on the Grid migration to Harvester. We have built a redundant architecture based on CERN...
Main Authors: | , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
EDP Sciences
2020-01-01
|
Series: | EPJ Web of Conferences |
Online Access: | https://www.epj-conferences.org/articles/epjconf/pdf/2020/21/epjconf_chep2020_03010.pdf |
id |
doaj-d4d016cb59cc40e0bc7ff3b4ddfd37e4 |
---|---|
record_format |
Article |
spelling |
doaj-d4d016cb59cc40e0bc7ff3b4ddfd37e42021-08-02T15:10:41ZengEDP SciencesEPJ Web of Conferences2100-014X2020-01-012450301010.1051/epjconf/202024503010epjconf_chep2020_03010Managing the ATLAS Grid through HarvesterBarreiro Megino Fernando Harald0Alekseev Aleksandr1Berghaus Frank2Cameron David3De Kaushik4Filipcic Andrej5Glushkov Ivan6Lin FaHui7Maeno Tadashi8Magini Nicolò9University of Texas at ArlingtonTomsk Polytechnic UniversityUniversity of VictoriaUniversity of OsloUniversity of Texas at ArlingtonJozef Stefan InstituteUniversity of Texas at ArlingtonUniversity of Texas at ArlingtonBrookhaven National LaboratoryIowa State UniversityATLAS Computing Management has identified the migration of all computing resources to Harvester, PanDA’s new workload submission engine, as a critical milestone for LHC Run 3 and 4. This contribution will focus on the Grid migration to Harvester. We have built a redundant architecture based on CERN IT’s common offerings (e.g. Openstack Virtual Machines and Database on Demand) to run the necessary Harvester and HTCondor services, capable of sustaining the load of O(1M) workers on the Grid per day. We have reviewed the ATLAS Grid region by region and moved as much possible away from blind worker submission, where multiple queues (e.g. single core, multi core, high memory) compete for resources on a site. Instead we have migrated towards more intelligent models that use information and priorities from the central PanDA workload management system and stream the right number of workers of each category to a unified queue while keeping late binding to the jobs. We will also describe our enhanced monitoring and analytics framework. Worker and job information is synchronized with minimal delays to a CERN IT provided ElasticSearch repository, where we can interact with dashboards to follow submission progress, discover site issues (e.g. broken Compute Elements) or spot empty workers. The result is a much more efficient usage of the Grid resources with smart, built-in monitoring of resources.https://www.epj-conferences.org/articles/epjconf/pdf/2020/21/epjconf_chep2020_03010.pdf |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Barreiro Megino Fernando Harald Alekseev Aleksandr Berghaus Frank Cameron David De Kaushik Filipcic Andrej Glushkov Ivan Lin FaHui Maeno Tadashi Magini Nicolò |
spellingShingle |
Barreiro Megino Fernando Harald Alekseev Aleksandr Berghaus Frank Cameron David De Kaushik Filipcic Andrej Glushkov Ivan Lin FaHui Maeno Tadashi Magini Nicolò Managing the ATLAS Grid through Harvester EPJ Web of Conferences |
author_facet |
Barreiro Megino Fernando Harald Alekseev Aleksandr Berghaus Frank Cameron David De Kaushik Filipcic Andrej Glushkov Ivan Lin FaHui Maeno Tadashi Magini Nicolò |
author_sort |
Barreiro Megino Fernando Harald |
title |
Managing the ATLAS Grid through Harvester |
title_short |
Managing the ATLAS Grid through Harvester |
title_full |
Managing the ATLAS Grid through Harvester |
title_fullStr |
Managing the ATLAS Grid through Harvester |
title_full_unstemmed |
Managing the ATLAS Grid through Harvester |
title_sort |
managing the atlas grid through harvester |
publisher |
EDP Sciences |
series |
EPJ Web of Conferences |
issn |
2100-014X |
publishDate |
2020-01-01 |
description |
ATLAS Computing Management has identified the migration of all computing resources to Harvester, PanDA’s new workload submission engine, as a critical milestone for LHC Run 3 and 4. This contribution will focus on the Grid migration to Harvester. We have built a redundant architecture based on CERN IT’s common offerings (e.g. Openstack Virtual Machines and Database on Demand) to run the necessary Harvester and HTCondor services, capable of sustaining the load of O(1M) workers on the Grid per day. We have reviewed the ATLAS Grid region by region and moved as much possible away from blind worker submission, where multiple queues (e.g. single core, multi core, high memory) compete for resources on a site. Instead we have migrated towards more intelligent models that use information and priorities from the central PanDA workload management system and stream the right number of workers of each category to a unified queue while keeping late binding to the jobs. We will also describe our enhanced monitoring and analytics framework. Worker and job information is synchronized with minimal delays to a CERN IT provided ElasticSearch repository, where we can interact with dashboards to follow submission progress, discover site issues (e.g. broken Compute Elements) or spot empty workers. The result is a much more efficient usage of the Grid resources with smart, built-in monitoring of resources. |
url |
https://www.epj-conferences.org/articles/epjconf/pdf/2020/21/epjconf_chep2020_03010.pdf |
work_keys_str_mv |
AT barreiromeginofernandoharald managingtheatlasgridthroughharvester AT alekseevaleksandr managingtheatlasgridthroughharvester AT berghausfrank managingtheatlasgridthroughharvester AT camerondavid managingtheatlasgridthroughharvester AT dekaushik managingtheatlasgridthroughharvester AT filipcicandrej managingtheatlasgridthroughharvester AT glushkovivan managingtheatlasgridthroughharvester AT linfahui managingtheatlasgridthroughharvester AT maenotadashi managingtheatlasgridthroughharvester AT magininicolo managingtheatlasgridthroughharvester |
_version_ |
1721230855207124992 |