RESILIENCE
Disaster recovery at MedTalk
The technical recovery approach for the MedTalk AI platform: recovery objectives, failure scenarios, and how recovery is tested.
Purpose and scope
This plan sets out the technical recovery approach for the MedTalk AI platform, hosted on Amazon Web Services in the Sydney (ap-southeast-2) region. It covers infrastructure, application, and database recovery, and is the technical execution layer supporting MedTalk's Business Continuity Management Policy.
Recovery objectives
| Objective | Target | Basis |
|---|---|---|
| Recovery Time Objective (RTO) | 4 hours | Time to restore core platform availability from a significant infrastructure failure |
| Recovery Point Objective (RPO) | 15 minutes | Maximum acceptable data loss window, based on continuous replication and automated backups |
Failure scenarios and recovery approach
- Availability Zone failure: MedTalk's AWS Sydney architecture spans multiple Availability Zones. A single Availability Zone failure is designed to be absorbed automatically, with workloads and data continuing to serve from the remaining zones without manual intervention.
- Region failure:In the event of a broader AWS Sydney region-level disruption, MedTalk's recovery runbook governs failover to a secondary recovery configuration, prioritising restoration of identity, core application services, and database access in that order, against the 4-hour RTO.
- Database recovery: Databases use automated, continuous backups with point-in-time recovery. Restoration is performed from the most recent consistent recovery point within the 15-minute RPO, verified for integrity before the application tier is reconnected.
- Application tier recovery: Application services run as containerised workloads on Amazon ECS Fargate, allowing rapid redeployment from version-controlled infrastructure and container definitions rather than manual server rebuilding.
Recovery sequence
Recovery follows a fixed sequence to restore service in a controlled, verifiable order.
| Step | Action | Owner |
|---|---|---|
| 1 | Declare incident and activate the crisis management process under the Business Continuity Management Policy | CISO / Incident Commander |
| 2 | Restore identity and access services (SSO integration) | MedTalk DevOps |
| 3 | Restore database services from the most recent verified recovery point | MedTalk DevOps |
| 4 | Restore application and AI processing tier | MedTalk DevOps |
| 5 | Verify end-to-end functionality before returning the service to clinicians | CISO |
| 6 | Notify customers of restoration and begin post-incident review | Incident Commander |
Backup standards
Daily full snapshots are encrypted using dedicated AWS Key Management Service keys, with continuous incremental backups supporting the 15-minute RPO. Backups are stored separately from production data to prevent a single failure or compromise from affecting both.
Recovery runbooks
Step-by-step recovery runbooks are maintained for each failure scenario in this plan, covering the specific commands, access requirements, and verification checks needed to execute each recovery step. Runbooks are version-controlled alongside MedTalk's infrastructure code and updated whenever the underlying architecture changes.
DR testing and validation
- Quarterly Restoration Testing: Backup restoration is tested quarterly to confirm that recovery points are valid and that the stated RPO is achievable in practice.
- Annual Failover Exercise: MedTalk commits to conducting a full disaster recovery exercise at least annually, simulating a significant infrastructure failure and validating recovery against the 4-hour RTO.
- Findings and Remediation: Gaps identified during testing are tracked to closure, with runbooks and this plan updated to reflect lessons learned before the next scheduled test.
Recovery governance
Disaster recovery governance sits with the Technology and Security Council established under MedTalk's Enterprise Governance Framework, with the Chief Information Security Officer accountable for maintaining this plan, the associated runbooks, and the testing schedule set out above.
Contact
Security & Compliance: support@medtalk.co
Legal / compliance: legal@medtalk.co