Cloud and DevOps services
that make releases routine.
Cloud setup, release pipelines and monitoring in your own cloud account, with the infrastructure and the pipeline kept as code in your repository.
One path to production.
The pipeline builds each change once and tests it. The same build goes to staging, then to production.
Repository
Application and infrastructure code, reviewed.
Tests
Each change built once and tested.
Staging
Set up from the same code as production.
Production
The same build, released from the pipeline only.
And a way back.
When a release goes wrong, an alert reaches a named person. The way back is rehearsed on staging: a rollback to the last good release or, where a data change can’t be undone, a forward fix.
- Release 419 goes out from the pipeline and the error rate rises above the alert threshold.
- An alert reaches a named person.
- A rollback brings back the last good release, 418, and the error rate falls.
- Release 420 with the fix takes the same path to production.
When teams call us
Four signs it is time to look at the setup, and what changes afterwards.
Only one person dares to deploy
Afterwards: Any engineer on the team can release from the pipeline.
Every release is a live experiment
Afterwards: Each change runs on staging before it reaches production.
Incidents are investigated in server logs
Afterwards: Investigation starts on dashboards; an alert points to the signal that changed.
The bill grows faster than usage
Afterwards: The bill is broken down by service and environment.
What we set up
The smallest setup that fits your product and team.
Where it runs
Staging and production in separate cloud accounts you own, named access for each person and secret values kept out of the code.
Your cloud accounts
Separate accountsStaging
- App
- Database
- Backups
Production
- App
- Database
- Backups
- Secrets manager
- Named access
How changes ship
Infrastructure as code reviewed by a second engineer, a pipeline with tests, staging first and a way back.
Infrastructure change
Staging first- +Queue for background jobs
- ~Web service: 2 → 3 instances
- −Old build server
- Reviewed by a second engineer
- Applied by the pipeline
How you keep track
Metrics, logs and traces in one place, alerts routed to a named person, a restore tested before go-live and the bill by service.
This month, by service
Production- Alerts to a named person
- Restore tested
IaaS, PaaS
or SaaS?
Most teams come to us with a platform already running. Three questions give a first read; the final choice comes after we have looked at what runs today.
Architecture
Deploys
Where it runs
PaaS: a managed platform
- A good fit when
- A monolith in the cloud, deployed weekly or daily.
- Watch out for
- Limits on configuration, and costs to recheck as volume grows.
- Bought as SaaS
- CI, monitoring, email and sign-in: parts every product needs and nobody should rebuild.
Two platforms we built,
and where they run.
Order a new setup.
Or move a live system.
A move adds one stage: a rehearsal on staging, then the switch to production.
- 01
Review
What runs today, and what breaks.
- 02
Plan
The target setup, and what is left out.
- 03
Build
Staging first, then production, as code.
- 04
Hand over
Runbook, alerts and a session with your team.
Before anything is built
You approve the written plan and its fixed price.
What stays with your team
The cloud account, the runbook and alert routing. Infrastructure and pipeline stay as code in your repository.
Want releases
to feel routine?
Tell us what runs today and what keeps going wrong. We will come back with questions and the smallest setup we would suggest; the price is fixed once the scope is written.
Discuss your setupFrequently Asked Questions
A fixed price for a written scope, agreed after a first call about what runs today and where you want to be. The price depends on how much of the current setup is documented, how many services and environments the work covers, whether a live system has to move and how much downtime the business can accept, your access and security rules, and the handover. Ongoing support after launch can be arranged separately.
Yes. A release pipeline, monitoring and alerts, backups or the move of one service can each be a separate project. We start with the part that causes the most trouble; the rest can wait or stay with your team.
Only when your services call for it. Kubernetes fits microservices and many deploys a day, in the cloud or in your own data centre. A monolith deployed weekly or daily usually runs well on a managed platform (PaaS), with less to look after. Most teams come to us with a platform already running, so we start from what runs today and tell you what to change.
You do. We work in your account with named access for each person, which you can revoke when the project ends. Infrastructure, pipeline and alert rules stay as code in your repository.
Yes. Ongoing support after launch can be arranged separately, with its scope agreed in writing.
Not as a promise. The review shows where the money goes, by service and environment, and which changes could lower the bill. You decide which of them are worth making.
AWS for the cloud, and Docker and Kubernetes for containers. The research data platform we built runs on AWS, and the marketplace seller platform runs in Docker and Kubernetes.
Yes. Cloud and DevOps work can be ordered on its own or together with custom software development or SaaS development. If you need the whole system reviewed first, start with a technical audit.
