Data Use Policy

[ COLIBRÍ // GOVERNANCE // DATA POLICY ]

HPC Storage & Data Stewardship Policy

The NMSU High Performance Computing (HPC) program provides data storage services to support computational research while promoting responsible data stewardship consistent with the FAIR Guiding Principles: Findable, Accessible, Interoperable, and Reusable.

POLICY SCOPE

This policy defines storage classes, allocation mechanisms, stewardship responsibilities, and data retention practices. It applies to all data stored on NMSU HPC-managed storage systems and all users granted access through:

  • Baseline institutional allocations, and/or
  • Project-specific or buy-in ("condo") storage arrangements

* APPLIES REGARDLESS OF FUNDING SOURCE OR COST-RECOVERY MECHANISM

// DEFINITIONS

Allocation Storage

Capacity granted to a project or PI through an application or renewal process, without direct monetary charge up to a defined baseline.

Condo (Buy-In) Storage

Capacity acquired through time-bound financial contribution, providing enhanced capacity/performance while subject to HPC governance.

Stewardship Record

Project-level metadata record describing stored data, ownership, access conditions, retention expectations, and reuse intent.

// CORE_STEWARDSHIP_PRINCIPLES
  1. FAIR is not synonymous with open or free access.
    Data may be access-restricted or embargoed while remaining FAIR, provided metadata and access conditions are clearly described and persist over time.
  2. Storage funding does not override stewardship obligations.
    All HPC storage (allocated or purchased) must meet institutional data stewardship requirements.
  3. Persistent metadata is required even when data files are removed or archived.
    FAIR requires metadata to remain accessible beyond the lifetime of the data itself.
// STORAGE_ARCHITECTURE

Storage Tiers & Intended Use

Storage Tier Intended Use & Characteristics Backup & Retention
Scratch Storage Active computations and intermediate result processing. High-speed I/O. Not backed up.
Subject to automated purges.
Project (Active) Storage Persistent storage for active research projects. Baseline allocation or condo share. Backed up to HPC standards.
Requires active stewardship record.
Archive Storage Inactive or completed projects. Cost-effective long-term retention. Long-term preservation.
Not optimized for compute I/O.
// FINANCIAL_MODEL

Storage Allocation & Buy-In Model

Most researchers are used to cloud storage models ($X per TB per month). In an academic HPC setting, monthly billing discourages co-locating data with compute resources and turns long-term research data into a recurring financial liability. Instead, Colibrí utilizes a baseline allocation + buy-in (shares) model.

MODEL PART 1

Baseline Allocation

  • Granted to each approved HPC project for active research use without direct fees.
  • Supported initially by NSF CC* contributed capacity shares.
  • Guaranteed commitments are calculated prior to share-based entitlement calculations.
  • Institutional baseline capacity is capped at ~1.5 PB to accommodate 5–10 year growth while preserving ~79% of total capacity for project buy-in shares.
MODEL PART 2

Shares (Buy-In) Storage

  • A one-time pre-paid purchase for persistent storage over ~5 years (no monthly recurring billing).
  • Each share provides a guaranteed minimum of ~20 TB plus a soft quota to accommodate temporary spikes.
  • Data remains persistent when grants expire; research teams work with HPC staff to renew, archive, or curate data at end-of-service.
  • Buy-in confers additional capacity/performance but remains fully subject to FAIR stewardship rules.
// FAIR_STEWARDSHIP

FAIR Stewardship Requirements

F

Findability

Projects must maintain a registered metadata record with PI details, persistent internal IDs, dataset descriptions, keywords, and associated publication/grant links.

A

Accessibility

Records must specify access terms (open, controlled, restricted), authentication requirements, and embargo periods. Metadata remains online even if files are archived.

I

Interoperability

Projects should adopt accepted file formats, document community vocabularies, and detail software, dependencies, and execution workflows needed for data interpretation.

R

Reusability

Records must detail reuse scope (private, collaborative, public), licensing terms, and provenance data. Grant expiration alone does not relieve reuse documentation requirements.

// LIFECYCLE_&_RESPONSIBILITIES

Data Retention & End-of-Life

  1. Planned Retention: Declared at allocation/buy-in time. May involve migration to archive storage, external repositories, or disciplinary data centers.
  2. Allocation Expiration: Data is migrated or dispositioned according to the approved plan. Stewardship metadata persists regardless of file disposition.
  3. Abandoned Data: Data lacking an active PI or stewardship contact may be archived, moved to institutional custody, or purged after metadata preservation.

Roles & Responsibilities

Principal Investigators (PIs)
  • Maintain accurate stewardship records
  • Comply with funding agency data management plans
  • Designate successor curators prior to departure/project close
HPC Program Staff
  • Provide storage infrastructure & retention workflows
  • Maintain metadata inventory systems
  • Enforce allocation, renewal, and purge policies consistently
QUESTIONS ABOUT STORAGE OR STEWARDSHIP?

To request storage shares, register dataset metadata, or consult on Data Management Plans (DMPs), contact the HPC team through ColdFront or submit a research computing support ticket.