skip to main content

High Energy Physics Seminar

Monday, November 30, 2020
10:00am to 11:00am
Add to Cal
Online Event
Towards a heterogeneous computing farm for the CMS High Level Trigger
Andrea Bocci, CERN,

The Compact Muon Solenoid experiment is planning substantial upgrades of its detectors and data acquisition system to fully exploit the physics reach of the High-Luminosity Large Hadron Collider. The higher luminosity, pileup, detector granularity and event readout rate present an unprecedented challenge to the real-time and offline event reconstruction, requiring a processing power that is orders of magnitude larger than today.

In recent years, industry and high-performance computing centres have successfully used heterogeneous computing platforms to achieve higher throughput and better energy efficiency, combining traditional processors with dedicated accelerators and matching each task to the most appropriate architecture. The CMS collaboration has decided to evaluate a heterogeneous computing farm for the upcoming Run 3 (2022-2024), with the goal to improve the physics reach of the experiment and gain the necessary expertise in view of a wider deployment during the High-Luminosity runs at the LHC.

The speaker will present the potential benefits of adopting a heterogeneous HLT farm for Run 3 and the HL-LHC Phase 2 upgrades; discuss the challenges encountered while developing the CMS physics reconstruction software (CMSSW) to take advantage of GPUs; and conclude with an overview of "performance portability" solutions being evaluated to exploit different kind of accelerator technologies.

Meeting ID: 880 2304 8308