Hosting and platforms¶
Where and how services run. Use the paved road so your team can focus on users rather than infrastructure.
Doctrine: 1. Platforms before projects. Built as products, 2. Standards and guardrails before exceptions, 3. Reuse before buy
Principles: 3. Maximise value, minimise waste, 1. Delivery-focused architecture
Who these guardrails apply to
The core department. Whether the hosting and platforms guardrails also apply to Defra's arm's length bodies is still to be confirmed.
Technology capability Enabling Platforms: application hosting and delivery platform.
GR-HOST-01 Use Defra's strategic delivery platform by default¶
Must New digital services are hosted on the Defra Core Delivery Platform (CDP) unless an exception has been agreed.
Why: CDP provides secure-by-default hosting, CI/CD, observability, secrets management and protective monitoring once, for everyone. Each team that builds its own platform recreates this at its own cost and risk.
How to meet it: In discovery or alpha, work with the Delivery Architecture team to decide whether CDP is right for your service - the expectation is that it will be - and engage the platform team. If CDP cannot meet a requirement (for example specialist compute, a SaaS product or a legacy migration), raise an exception early and tell the platform team - the gap may be something they should solve for everyone.
In the Defra Digital Service Manual: Core Delivery Platform. The manual says a service not on CDP is managed as an exception through the Delivery Architecture team's governance process; how that relates to exceptions on this site is still being agreed.
GR-HOST-02 Public cloud first¶
Should Where a service cannot use a strategic platform, it is hosted in a Defra-managed public cloud tenancy. New on-premises hosting is not permitted.
Why: Government Cloud First policy and Defra's data centre exit.
GR-HOST-03 Everything as code¶
Should Infrastructure, configuration, pipelines and policies are defined as code, version-controlled and deployed through automated pipelines. No manual changes to production.
Why: Repeatable, reviewable, recoverable environments. Manual changes cause drift and incidents.
How to meet it: Use the platform's provided templates. Where you manage your own infrastructure, use a declarative tool such as Terraform and run drift detection.
GR-HOST-04 Use managed services before self-managed¶
Should Prefer managed cloud services (databases, queues, storage) over running your own on virtual machines.
Why: Managed services remove patching and much operational toil.
GR-HOST-05 Consistent, disposable environments¶
Should Have at least development, test and production environments, created from the same code, with production data never copied to lower environments unless anonymised.
GR-HOST-06 Host data in the UK¶
Should Data classified OFFICIAL is held in UK regions unless an assessed and approved exception exists.
Why: Data protection, sovereignty and Defra's information risk appetite.
GR-HOST-07 Design for the resilience the service needs¶
Should Agree recovery time and recovery point objectives with the service owner, design to them across availability zones, and test recovery at least once a year.
Why: Some Defra services, such as flood warnings and disease control, are critical during emergencies - exactly when infrastructure is under stress.