Skip to content
2026-07-01

Infrastructure Workstream

The Infrastructure Workstream covers setting up the secure network area, installing the query retrieval software, and connecting it to your OMOP data and the Cohort Discovery Service.

Responsible party

The Data Custodian is responsible for all infrastructure steps. The HDR UK Technology Team is available to answer questions but will not have direct access to your systems.

Steps overview

Step Description
I1 Set up a Secure Network Area
I2 Install Query Retrieving / Running Software
I3 Connect OMOP Data to Query Retrieving Software

I1 — Setting up a Secure Network Area

Description

Data Custodians must provide a secure environment within their network to house and process their pseudonymised OMOP data. The ETL tool (D4) will load data into this area, and the query retrieval software will be installed here so that queries can be performed.

Requirements:

  • Isolated from environments containing identifiable data
  • Contains only: (1) the pseudonymised OMOP database, and (2) the query retrieval software
  • Supports outbound HTTPS (port 443) only — no inbound firewall rules required
  • Supports OCI-compliant container runtime (Docker, Podman, or Kubernetes)

A dedicated VM is the most common deployment model. It provides clear separation between the pseudonymised OMOP data zone and other systems.

Cloud-based deployments (AWS, Azure, GCP) are fully supported. The secure network area can be implemented as a cloud VPC or subnet with appropriate network controls.

On-premises deployments are supported. The secure network area should be a dedicated VLAN or subnet with outbound-only firewall rules.

Governance preference

Our experience is that clear separation of data zones provides additional comfort to Data Governance panels. We recommend a distinct and minimal deployment model to simplify IG reviews.

Bunny infrastructure specification

Requirement Minimum Specification
CPU 2 vCPUs
Memory (RAM) 4 GB
Container Runtime OCI-compliant (Docker, Podman, or Kubernetes)
Network Outbound HTTPS (port 443) only
Database PostgreSQL 14–18 · SQL Server 2019 or 2022 · DuckDB 1.x · Snowflake

Requirements may vary depending on workload and deployment environment.

See the Bunny user guide or contact the University of Nottingham for optimal environment recommendations.

I2 — Install Query Retrieving / Running Software

This software routinely polls the Cohort Discovery API for tasks and runs them against your OMOP database.

Bunny is the recommended open-source option.

  • Developed by the University of Nottingham
  • Outbound-only requests
  • Supports obfuscation (low-count suppression and rounding)
  • Can participate in federated networks via Hutch Relay
  • Container images available: ghcr.io/health-informatics-uon/hutch/bunny:latest

Two instances required

Bunny requires two instances per OMOP database — one for availability queries (type a) and one for distribution jobs (type b). See the Two-Instance Pattern.

For connecting Bunny to the Cohort Discovery Service, see Connecting Bunny and Appendix 2 — Step-by-Step Setup.

Bunny user guide: health-informatics-uon.github.io/hutch/bunny

The Cohort Discovery Service is compatible with BC|INSIGHT as a query retrieval tool.

As this is a proprietary tool, contact BC Platforms directly for: - Installation instructions - Contract options and pricing - Technical support

Data Custodians are welcome to develop their own query retrieval tool.

Requirements: - Must meet the API standards set out in the Swagger documentation - Contact the HDR UK Technology Team to discuss this option before starting development

I3 — Connect OMOP Data to Query Retrieving Software

Description

Once the OMOP database has been created and stored, it needs to be connected to the query retrieval software.

  1. Upload the OMOP database to the same network space as the query retrieval software
  2. Connect the database to the query retrieval software (if not already connected)

  3. Ensure recommended indexes are in place on your OMOP tables (for Bunny):

    Index Table
    idx_person_id Person
    idx_concept_concept_id Concept
    idx_condition_concept_id_1 Condition Occurrence
    idx_procedure_concept_id_1 Procedure Occurrence
    idx_observation_concept_id_1 Observation
    idx_measurement_concept_id_1 Measurement

    See hutch.health/bunny/deployment/requirements for the full index list.

  4. Connect the query retrieval software to Cohort Discovery — follow the steps in Connecting Bunny

  5. Test that queries can run from the Cohort Discovery Service

Test with your draft Collection

While your Collection is in Draft status, you can run test queries visible only to you. See Step 8 of the Bunny setup guide.

Monitoring your Collection

We recommend setting up internal monitoring to detect when your Collection goes offline. The most practical approach is to monitor your Bunny instance health and logs — Bunny's own documentation covers available logging and health-check options. You can also check Collection status at any time in the Cohort Discovery admin panel; a Collection that has gone offline will show as Offline or Suspended in your Collections list. Contact us at gateway@hdruk.ac.uk if you need support diagnosing a persistent issue.