Kory Draughn

Chief Technologist, iRODS Consortium

November 16-21, 2025

Supercomputing 2025

St. Louis, MO

iRODS Policy-Based Data Management

iRODS has always been a programmable virtual filesystem, providing rule-based operations on data files stored on disparate physical storage systems, both local and remote.

It stores both:

• standard system metadata and
• additional arbitrary user-defined metadata

And this all lives in a relational database, available via SQL.

iRODS Timeseries

But...

iRODS does not provide historical information about "what used to be".

The way to know "what used to be", is to have asked, and to have saved it.

iRODS Timeseries

Best practice:

• Regularly query the database for 'metrics'
• Write them down
• Later, query the metrics, and draw some graphs

sql_exporter → prometheus → grafana

• https://github.com/burningalchemist/sql_exporter
• https://github.com/prometheus/prometheus
• https://github.com/grafana/grafana

iRODS Timeseries

This is now implemented in the irods_demo repository

• https://github.com/irods/irods_demo

iRODS Dashboarding

iRODS Dashboarding

poor man's placeholders from bison / irods_demo

iRODS Chargebacks

Captures transfers in and out of the namespace

• Demonstrated via iRODS Rule Language
• Easily hardened, optimized to C++

This provides administrators with comprehensive network usage information, suitable for scripting and automating chargebacks in a service-oriented environment.

Savings through Storage Tiering

iRODS also now tracks access_time for every replica in the system.

Leads to a straightforward implementation of storage tiering

• https://github.com/irods/irods_capability_storage_tiering

This allows for direct measurement of savings from keeping 'cold' data on cheaper storage.

Storage Tiering

Impact

iRODS Storage Tiering Savings Calculator

Thank you!

Booth #4424