CERN Accelerating science

Schematic view of the {\submit} components and relations to other local and global clusters. Local SubMIT resources support batch computing via Slurm as highlighted in blue. Global clusters for batch computing via HTCondor highlighted in yellow are the CMS tier 2 (T2) and tier 3 centers (T3) at MIT, the CMS global computing pool, and the open science grid (OSG). The Lattice QCD (LQCD) cluster at MIT is also accessible. Different storage systems are colored in green.

(left) Total number of registered users since opening the service to the department. (right) Number of distinct users that used {\submit} each week from December 2024 to March 2025. The users are categorized by academic position.

(left) Total number of registered users since opening the service to the department. (right) Number of distinct users that used {\submit} each week from December 2024 to March 2025. The users are categorized by academic position.

Population of research users categorized by department (left) and division within the physics department (right). The category ``other'' also contains the users that are external to MIT. Indicated in brackets are the number of users per category. If a user is associated to multiple categories the user is counted in fractions to each contributing category.

Type of user interaction with the system in terms of software (left) and hardware (right).

Example of monitoring images for the total number of jobs as a function of time submitted via {\slurm} for a time window of a day (left) or {\htc} for a time window of a week (right). The colors indicate the status of a job, and if running, on which partition or site it is being executed.

Example of monitoring images for the total number of jobs as a function of time submitted via {\slurm} for a time window of a day (left) or {\htc} for a time window of a week (right). The colors indicate the status of a job, and if running, on which partition or site it is being executed.

Left: Performance comparison across storage system: time spent to read from different input sources. Right: Performance comparison across computing configurations: time spent to execute the task locally on a node or remote on multi-nodes.

Left: Performance comparison across storage system: time spent to read from different input sources. Right: Performance comparison across computing configurations: time spent to execute the task locally on a node or remote on multi-nodes.

The Hrare analysis workflow. The computing resources that are used to analyse the data are shown above each step while the approximate storage space is shown below in purple. The dashed boxes represent different stages of the analysis that require different software as shown in the bottom. The approximate time required to process files for each stage is shown at the top in blue.