Evolution and experience with the ATLAS Simulation at Point1 Project

S Ballestrero,F Brasolin,A Di Girolamo,J Schovancova,A Zaytsev,T Kouba,F Wang,D Fazio,M S Twomey,C J Lee,D A Scannicchio

doi:10.1088/1742-6596/898/8/082012

Abstract

The Simulation at Point1 project is successfully running standard ATLAS simulation jobs on the TDAQ HLT resources. The pool of available resources changes dynamically, therefore we need to be very effective in exploiting the available computing cycles. We present our experience with using the Event Service that provides the event-level granularity of computations. We show the design decisions and overhead time related to the usage of the Event Service. The improved utilization of the resources is also presented with the recent development in monitoring, automatic alerting, deployment and GUI.

Highlights

During the LHC [1] Long shutdown 1 (LS1) period, the Simulation at Point1 (Sim@P1) project utilized the TDAQ HLT [2] (Trigger and Data Acquisition system, High Level Trigger) resources very well in a continuous manner
LHC Run 2 period presented slightly different challenge as the Simulation at Point1 project successfully continued in using the resources reserved for Sim@P1 and we focused on better utilization of the resources given to Sim@P1 during the periods when the resources are not needed for TDAQ HLT computing
The chart shows that Sim@P1 contributed more than 12% of all the events simulated for ATLAS since March 2015

Summary

Introduction

During the LHC [1] Long shutdown 1 (LS1) period, the Simulation at Point (Sim@P1) project utilized the TDAQ HLT [2] (Trigger and Data Acquisition system, High Level Trigger) resources very well in a continuous manner. LHC Run 2 period presented slightly different challenge as the Simulation at Point project successfully continued in using the resources reserved for Sim@P1 and we focused on better utilization of the resources given to Sim@P1 during the periods when the resources are not needed for TDAQ HLT computing. Such periods are usually Machine Development (MD) or Technical Stop (TS).

Event Service pilot test

Automated switcher: “red button”

Monitoring developments

Dynamic partitioning and Cgroups

Findings

Summary