Computing demands for large scientific experiments, such as the CMS experiment at the CERN LHC, will increase dramatically in the next decades. To complement the future performance increases of software running on central processing units (CPUs), explorations of coprocessor usage in data processing hold great potential and interest. Coprocessors are a class of computer processors that supplement CPUs, often improving the execution of certain functions due to architectural design choices. We explore the approach of Services for Optimized Network Inference on Coprocessors (SONIC) and study the deployment of this as-a-service approach in large-scale data processing. In the studies, we take a data processing workflow of the CMS experiment and run the main workflow on CPUs, while offloading several machine learning (ML) inference tasks onto either remote or local coprocessors, specifically graphics processing units (GPUs). With experiments performed at Google Cloud, the Purdue Tier-2 computing center, and combinations of the two, we demonstrate the acceleration of these ML algorithms individually on coprocessors and the corresponding throughput improvement for the entire workflow. This approach can be easily generalized to different types of coprocessors and deployed on local CPUs without decreasing the throughput performance. We emphasize that the SONIC approach enables high coprocessor usage and enables the portability to run workflows on different types of coprocessors.
Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher.
Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?
Some links on this page may take you to non-federal websites. Their policies may differ from this site.
-
Abstract Free, publicly-accessible full text available December 1, 2025 -
Free, publicly-accessible full text available November 1, 2025
-
Free, publicly-accessible full text available October 1, 2025
-
Free, publicly-accessible full text available October 1, 2025
-
Free, publicly-accessible full text available October 1, 2025
-
Abstract The operation and performance of the Compact Muon Solenoid (CMS) electromagnetic calorimeter (ECAL) are presented, based on data collected in pp collisions at √
s =13 TeV at the CERN LHC, in the years from 2015 to 2018 (LHC Run 2), corresponding to an integrated luminosity of 151 fb-1. The CMS ECAL is a scintillating lead-tungstate crystal calorimeter, with a silicon strip preshower detector in the forward region that provides precise measurements of the energy and the time-of-arrival of electrons and photons. The successful operation of the ECAL is crucial for a broad range of physics goals, ranging from observing the Higgs boson and measuring its properties, to other standard model measurements and searches for new phenomena. Precise calibration, alignment, and monitoring of the ECAL response are important ingredients to achieve these goals. To face the challenges posed by the higher luminosity, which characterized the operation of the LHC in Run 2, the procedures established during the 2011–2012 run of the LHC have been revisited and new methods have been developed for the energy measurement and for the ECAL calibration. The energy resolution of the calorimeter, for electrons from Z boson decays reaching the ECAL without significant loss of energy by bremsstrahlung, was better than 1.8%, 3.0%, and 4.5% in the |η | intervals [0.0,0.8], [0.8,1.5], [1.5, 2.5], respectively. This resulting performance is similar to that achieved during Run 1 in 2011–2012, in spite of the more severe running conditions.Free, publicly-accessible full text available September 1, 2025 -
A search for collective effects inside jets produced in proton-proton collisions is performed via correlation measurements of charged particles using the CMS detector at the CERN LHC. The analysis uses data collected at a center-of-mass energy of, corresponding to an integrated luminosity of. Jets are reconstructed with the anti-algorithm with a distance parameter of 0.8 and are required to have transverse momentum greater than 550 GeV and pseudorapidity. Two-particle correlations among the charged particles within the jets are studied as functions of the particles’ azimuthal angle and pseudorapidity separations (and) in a jet coordinate basis, where particles’,are defined relative to the direction of the jet. The correlation functions are studied in classes of in-jet charged-particle multiplicity up to. Fourier harmonics are extracted from long-range azimuthal correlation functions to characterize azimuthal anisotropy for. For low-jets, the long-range elliptic anisotropic harmonic,, is observed to decrease with. This trend is well described by Monte Carlo event generators. However, a rising trend foremerges at, hinting at a possible onset of collective behavior, which is not reproduced by the models tested. This observation yields new insights into the dynamics of jet evolution in the vacuum.
© 2024 CERN, for the CMS Collaboration 2024 CERN Free, publicly-accessible full text available September 1, 2025 -
A search is described for the production of a pair of bottom-type vectorlike quarks (VLQs) with mass greater than 1000 GeV. EachVLQ decays into aquark and a Higgs boson, aquark and aboson, or aquark and aboson. This analysis considers both fully hadronic final states and those containing a charged lepton pair from aboson decay. The products of theboson decay and of the hadronicorboson decays can be resolved as two distinct jets or merged into a single jet, so the final states are classified by the number of reconstructed jets. The analysis uses data corresponding to an integrated luminosity ofcollected in proton-proton collisions atwith the CMS detector at the LHC from 2016 to 2018. No excess over the expected background is observed. Lower limits are set on theVLQ mass at the 95% confidence level. These depend on theVLQ branching fractions and are 1570 and 1540 GeV for 100%and 100%, respectively. In most cases, the mass limits obtained exceed previous limits by at least 100 GeV.
© 2024 CERN, for the CMS Collaboration 2024 CERN Free, publicly-accessible full text available September 1, 2025 -
Abstract The performance of muon tracking, identification, triggering, momentum resolution, and momentum scale has been studied with the CMS detector at the LHC using data collected at √(
s NN) = 5.02 TeV in proton-proton (pp) and lead-lead (PbPb) collisions in 2017 and 2018, respectively, and at √(s NN) = 8.16 TeV in proton-lead (pPb) collisions in 2016. Muon efficiencies, momentum resolutions, and momentum scales are compared by focusing on how the muon reconstruction performance varies from relatively small occupancy pp collisions to the larger occupancies of pPb collisions and, finally, to the highest track multiplicity PbPb collisions. We find the efficiencies of muon tracking, identification, and triggering to be above 90% throughout most of the track multiplicity range. The momentum resolution and scale are unaffected by the detector occupancy. The excellent muon reconstruction of the CMS detector enables precision studies across all available collision systems.Free, publicly-accessible full text available September 1, 2025 -
A bstract A search for a new charged particle X with mass between 0.3 and 2.0 TeV decaying to a W boson and a photon is presented, using proton-proton collision data at a center-of-mass energy of 13 TeV, collected by the CMS experiment and corresponding to an integrated luminosity of 138 fb
− 1. Particle X has electric charge ±1 and is assumed to have spin 0. The search is performed using the electron and muon decays of the W boson. No significant excess above the predicted background is observed. The upper limit at 95% confidence level on the product of the production cross section of the X and its branching fraction to a W boson and a photon is found to be 94 (137) fb for a 0.3 TeV resonance and 0.75 (0.81) fb for a 2.0 TeV resonance, for an X width-to-mass ratio of 0.01% (5%). This search presents the most stringent constraints to date on the existence of such resonances across the probed mass range. A statistical combination with an earlier study based on the hadronic decay mode of the W boson is also performed, and the upper limit at 95% confidence level for a 2.0 TeV resonance is reduced to 0.50 (0.63) fb for an X width-to-mass ratio of 0.01% (5%).Free, publicly-accessible full text available September 1, 2025