Cloud, Hosting & Deployment

Most of the businesses we work with do not need a cloud strategy. They need deployments that do not require someone to be awake, backups that have been restored at least once, and a server that is patched. Those three things prevent more downtime than any architecture diagram.

Deployment you can repeat

Releases should be boring: the same steps every time, reversible, and runnable by someone who did not write the code. We set up pipelines that run migrations safely, keep environment configuration out of the repository, and let you roll back without a database restore. If deploying is frightening, it happens rarely, and rare deployments are large and risky ones.

Hosting sized to the actual load

Plenty of production systems run comfortably on a single well-configured server, and we will say so rather than sell you an orchestration layer to administer. When load genuinely warrants it we scale out, but the honest answer for most small and mid-sized operations is a smaller bill than they expect.

Third-party and open-source platforms

Not everything should be built. Where a mature open-source system already solves the problem, deploying and adapting it is faster and cheaper than writing another one. We deploy, configure, customise, migrate data into and support platforms of that kind for clients — Snipe-IT for IT asset and inventory tracking is one we have set up and run more than once.

To be clear about what that is: Snipe-IT is third-party open-source software published under the AGPL, not something we wrote. The work we do is the deployment, the data migration, the integration with your existing systems, the customisation and the ongoing support — which is the part that decides whether it succeeds in your business.

Migrations without a shutdown

Moving a live system between hosts is mostly planning: what has to be read-only and for how long, how data is reconciled afterwards, and what the rollback is if the new environment misbehaves at nine on a Monday. We have moved systems that could not stop, and the plan is the deliverable.

Monitoring worth having

Uptime checks, error tracking and alerts that go to a person who can act. Alerts nobody responds to are worse than none, because they teach a team to ignore the channel.