Data Management Plan Research Example

Standard Operating Procedure: Research Data Management Plan (DMP) Architecture & Execution

1. Document Control Block

• Document ID: SOP-TR-ENG-409
• Effective Date: October 24, 2023
• Version: 2.1.0
• Review Cadence: Annual
• Classification: Institutional Operations / Engineering Standards

---

2. Executive Summary & Purpose

This Standard Operating Procedure (SOP) defines the institutional engineering standard for designing, reviewing, and executing a Research Data Management Plan (DMP) within Template Registry. The purpose is to ensure absolute data integrity, metadata compliance, reproducibility, and security across all research lifecycles. Adherence to this protocol mitigates data loss vectors, ensures institutional compliance with funding body mandates (e.g., NSF, NIH, Horizon Europe), and secures the long-term archival viability of analytical assets.

---

3. Scope & Prerequisites

3.1 Scope

This procedure applies to all research systems, engineering squads, data stewards, and principal investigators operating under the Template Registry technical umbrella. It covers data collection, active processing, storage architecture, sharing protocols, and long-term preservation.

3.2 Prerequisites & Toolchain

• Access Control: Active Directory administrative privileges; RBAC clearance for Tier-3 sensitive datasets.
• Software/Toolchain:
  • DMPTool (or institutional equivalent) for schema generation.
  • Git / GitHub Enterprise for version-controlled documentation.
  • Apache Parquet / HDF5 for structured analytical outputs.
  • MinIO / AWS S3 for object storage tiering.
• PPE (Policy, Protocol, & Ethics): IRB protocol approval (if human subjects data is involved); Institutional Information Security Training certification.

---

4. Roles & Responsibilities

---

5. Step-by-Step Procedure

Phase 1: Data Discovery & Classification

• 1.1 Identify and inventory all anticipated data inputs, intermediate transformations, and final outputs.
• 1.2 Classify data sensitivity tier (Tier 1: Public, Tier 2: Internal/Proprietary, Tier 3: Restricted/PII/PHI).
• 1.3 Document data volume projections, growth rates, and required throughput parameters for active processing pipelines.

Phase 2: Storage Architecture & Infrastructure Provisioning

• 2.1 Establish primary storage buckets utilizing immutable object storage (e.g., AWS S3 Object Lock enabled in compliance mode).
• 2.2 Configure automated lifecycle policies shifting raw data from hot storage to cold archival tiers (e.g., Glacier Deep Archive) after 90 days of pipeline inactivity.
• 2.3 Implement synchronous multi-region replication for Tier-2 and Tier-3 assets to guarantee business continuity.

Phase 3: Metadata, Documentation, & Schema Enforcement

• 3.1 Define machine-readable schemas using JSON Schema or Apache Avro for all structured datasets.
• 3.2 Enforce Dublin Core or DDI metadata standards for dataset cataloging within the Template Registry central repository.
• 3.3 Maintain a living README.md and data dictionary in the project repository detailing variable definitions, unit of measurement, and known error bounds.

Phase 4: Security, Access Control, & Compliance

• 4.1 Enforce Principle of Least Privilege (PoLP) via IAM policies and role-based access control.
• 4.2 Apply cryptographic erasure and AES-256 encryption at rest; TLS 1.3 for all data in transit.
• 4.3 Verify IRB and ethics board sign-offs if handling human subject data, ensuring pseudonymization or hardware-level tokenization.

Phase 5: Archival, Sharing, & Long-Term Preservation

• 5.1 Prepare datasets for public repository deposition (e.g., Dryad, Zenodo, or institutional dataverse) upon project completion or publication.
• 5.2 Mint persistent identifiers (DOIs) for all released analytical assets to ensure citation traceability.
• 5.3 Execute a final bit-rot audit using SHA-256 checksum verification prior to long-term cold storage handoff.

---

6. Quality Assurance & Pro-Tips

6.1 Best Practices

• Automation First: Treat the DMP as code. Store the DMP configuration in version control alongside the pipeline orchestration scripts (e.g., Airflow DAGs).
• Early Security Review: Engage the Information Security Officer during Phase 1 to prevent costly architectural refactoring downstream.

6.2 Common Pitfalls

• The "Write-Only" Data Lake: Storing raw data without metadata or schema definitions, rendering it unrecoverable after 6 months.
• Neglecting Growth Metrics: Failing to account for exponential data expansion in active simulation runs, leading to pipeline throttling.

6.3 Metric Thresholds

• Checksum Verification Success Rate: 100% pass rate required across all cold-storage transfers.
• RTO / RPO: Recovery Time Objective (RTO) < 4 hours; Recovery Point Objective (RPO) < 1 hour for active analytical pipelines.

---

7. Frequently Asked Questions

Q1: What happens if the data classification changes mid-project (e.g., from Tier 2 to Tier 3)?

A: Halt active pipeline processing immediately. Notify the Information Security Officer within 2 hours. Re-provision the storage namespace under high-security encryption boundaries, execute an audit of current access logs, and revoke unauthorized IAM roles before resuming operations.

Q2: How should unstructured data (e.g., raw binary logs, unstructured text) be handled under this DMP framework?

A: Unstructured data must be ingested through standard normalization pipelines to extract structural metadata. If raw preservation is strictly required, store the binaries within immutable object containers accompanied by a companion sidecar JSON file detailing parsing instructions and provenance.