XDMoD

XDMoD
Name	XDMoD
Developer	University of Illinois National Center for Supercomputing Applications; Texas Advanced Computing Center; Indiana University; Ohio Supercomputer Center
Released	2009
Latest release	(varies)
Programming language	Python; PHP; JavaScript; MySQL
Operating system	Linux; CentOS; Red Hat Enterprise Linux
License	BSD-style; open source

XDMoD

XDMoD is a scalable performance monitoring and management suite for high-performance computing clusters and supercomputers. It provides historical accounting, real-time health checks, performance analysis, and chargeback reporting used by academic centers, national laboratories, and commercial providers. The project is employed alongside centers such as the National Center for Supercomputing Applications, Texas Advanced Computing Center, Oak Ridge National Laboratory, and the Pittsburgh Supercomputing Center.

Overview

XDMoD originated as a collaboration among academic and national research organizations including the National Science Foundation, University of Illinois, Texas Advanced Computing Center, Indiana University, and Ohio Supercomputer Center. It addresses operational needs similar to those tackled by projects at Lawrence Berkeley National Laboratory, Los Alamos National Laboratory, Argonne National Laboratory, and Sandia National Laboratories. The toolkit complements workload managers and schedulers like Slurm, PBS Professional, LSF, and Grid Engine, and interoperates with monitoring frameworks such as Nagios, Zabbix, Ganglia, and Prometheus. Institutions deploying XDMoD range from the European Centre for Medium-Range Weather Forecasts to national facilities such as PRACE centers, NERSC, and HPE, and it integrates into ecosystems involving vendors like IBM, Cray, Dell, and Lenovo.

Architecture and Components

XDMoD’s architecture includes data collectors, a relational database backend, web front end, analysis engines, and alerting subsystems. Collectors ingest job accounting records from Slurm, PBS, LSF, and SGE, telemetry from IPMI, SNMP, and Prometheus exporters, and filesystem metrics from Lustre, GPFS, and BeeGFS deployments. The database layer uses MySQL or MariaDB similar to backends used by CERN, Fermilab, and the European Synchrotron Radiation Facility. The web application is built with PHP and JavaScript frameworks akin to Drupal and Django-based portals at NASA and ESA science centers. Analysis modules produce summaries comparable to reports generated by Berkeley Lab’s perfSONAR and monitoring suites used at the Max Planck Society, RIKEN, and the French Alternative Energies and Atomic Energy Commission.

Features and Functionality

XDMoD delivers accounting reports, job-level performance metrics, system-level utilization dashboards, and chargeback invoicing. It supports performance tools and libraries such as PAPI, TAU, HPCToolkit, and Intel VTune for in-depth profiling, and links with MPI implementations including OpenMPI and MPICH used by projects at the European Grid Infrastructure and Jülich Supercomputing Centre. Energy and power monitoring integrates with IPMI and vendor power APIs similar to practices at Oak Ridge and Barcelona Supercomputing Center. Users can generate allocation reports used by NSF XSEDE allocations, DOE INCITE awardees, ERC-funded teams, and national research councils. Alerting and anomaly detection are comparable to thresholds used at CERN’s Large Hadron Collider computing facilities and genome centers like the Wellcome Sanger Institute.

Deployment and Integration

Deployments typically occur on Red Hat Enterprise Linux or CentOS systems in environments managed by administrative groups such as university research computing departments, national laboratory operations teams, and corporate HPC service units. Integration patterns mirror those at the San Diego Supercomputer Center, Pittsburgh Supercomputing Center, Texas Advanced Computing Center, and Compute Canada sites, connecting to identity providers like LDAP, Kerberos, Shibboleth, and OAuth used by the European Middleware Initiative and Internet2. Containerization strategies involve Docker and Kubernetes as adopted by Google, Microsoft Azure, Amazon Web Services, and OpenStack private clouds. Integration with job schedulers, storage systems, and telemetry aligns with workflows at CERN, SLAC National Accelerator Laboratory, and the Broad Institute.

Use Cases and Applications

Administrators use XDMoD for capacity planning, trend analysis, and allocation accounting analogous to capacity programs at NASA Ames Research Center and the Jet Propulsion Laboratory. Research groups employ it to profile scientific codes common in fields served by the National Center for Atmospheric Research, the Max Planck Institutes, and the Lawrence Livermore National Laboratory. Operational teams at climate modeling centers, genomics centers, and computational chemistry groups use XDMoD outputs in conjunction with tools like Ansible, Puppet, and SaltStack for automation. Funding agencies and university offices rely on its chargeback and accounting reports to audit resource usage for grants from the National Institutes of Health, European Research Council, and national science foundations.

Development and Community

The project has an open-source governance model with contributors from academic institutions, national labs, and commercial partners, reflecting collaborative communities similar to Apache Software Foundation projects and the Linux kernel community. Development follows issue tracking and code review workflows used at GitHub-hosted projects by institutions such as MIT, Stanford, and UC Berkeley. Training materials, workshops, and user conferences align with community activities at Supercomputing Conference, ISC High Performance, PEARC, and the International Supercomputing Conference. Collaborative integrations have been contributed by teams at IBM Research, Intel, NVIDIA, HPE, and Mellanox.

Security and Privacy

Operational security practices for XDMoD deployments align with standards and compliance frameworks used by U.S. Department of Energy laboratories, European data protection regimes, and institutional IT security offices at Harvard, Oxford, and Cambridge. Authentication and authorization use enterprise systems like LDAP, Kerberos, and Shibboleth; audit logging and role-based access control mirror procedures at CERN and national research infrastructures. Privacy considerations for researcher attribution and usage reporting are managed in ways comparable to policies at the Wellcome Trust Sanger Institute and EMBL-EBI, ensuring sensitive project identifiers and personally identifiable information are handled per institutional policy.

Category:Supercomputing