PanDA: Evolution and Recent Trends in LHC Computing

详细信息    查看全文

• 作者：Fernando Barreiro Megino^a ; ; Kaushik De^a ; Jose Caballero^b ; John Hover^b ; Alexei Klimentov^b ; Tadashi Maeno^b ; Paul Nilsson^b ; Danila Oleynik^a ; Sergey Padolski^b ; Sergey Panitkin^b ; Artem Petrosyan^c ; Torre Wenaus^b ; on behalf of the ATLAS collaboration
• 关键词：Workload Management System ; ATLAS ; Grid Computing ; Cloud Computing ; HPC
• 刊名：Procedia Computer Science
• 出版年：2015
• 出版时间：2015
• 年：2015
• 卷：66
• 期：Complete
• 页码：439-447
• 全文大小：1932 K

文摘

The Large Hadron Collider(LHC) is the world's largest and most powerful machine. It started operating in 2009 with a scientific program foreseen to extend over the next coming decades at increasing energies and luminosities to maximise the discovery potential. During Run1 (2009- 2013), the Worldwide LHC Computing Grid (WLCG) successfully delivered all the necessary computing resources, which made the discovery of the Higgs Boson possible. Looking ahead, it is forecasted that increased luminosities will extrapolate to a multiplicity in the storage and processing costs, which is not reflected in a corresponding funding growth of the WLCG. ATLAS, one of the four experiments at the LHC, is therefore leading an upgrade program to evolve their software and computing model to make the best possible usage of available resources, and also leverage on upcoming state of the art computing paradigms that could make important resource contributions.

These proceedings will give an insight into the accompanying work in PanDA, ATLAS’ workload management system. PanDA has implemented event level bookkeeping and dynamic generation of jobs with tailored lengths, in order to integrate and optimise the usage of oppor- tunistic resources, e.g. Cloud Computing or High Performance Computing (HPC). In conjunc- tion, the Event Service has been developed as a way to manage fine grained jobs and its outputs. Usage examples on some of the leading commercial and research infrastructures will be given. In addition, we will describe the work on further exploiting the current network capabilities by allowing remote data access and reducing regional boundaries.