Multi-cloud migration for travel-tech

Traffic was seasonal and spiky, the architecture was over-provisioned for the peaks, and a single provider outage could take the whole platform offline. Cost and reliability had become the same conversation.

The challenge

The platform ran entirely on one cloud, sized for its busiest days and paying for that capacity around the clock. Deployments were manual and nervous, infrastructure lived in the console rather than in code, and there was little real visibility into where money and latency were actually going. When the provider had a bad day, so did every customer.

The CTO wanted three outcomes that usually pull against each other: lower spend, higher reliability, and no disruption to a live, revenue-generating service during the move.

Reliability and cost had become the same problem — and both pointed at the architecture.

What we did

We began by putting the whole estate into code and instrumenting it, so decisions were driven by data rather than guesswork. Then we designed a multi-cloud target across AWS and GCP, and migrated workload by workload behind a rollback plan, keeping the live service untouched until each piece was proven.

• Codified the entire infrastructure in Terraform — reproducible, reviewable, and portable across providers.
• Containerised and orchestrated workloads on Kubernetes so they run the same way on AWS and GCP.
• Built CI/CD pipelines that turned manual, risky releases into automated, repeatable deploys.
• Added full observability — metrics, logs and tracing — to surface the over-provisioning and right-size every service.
• Migrated with a phased, zero-downtime cutover and a tested fallback at each step.

"We moved to a multi-cloud setup with zero downtime and cut our monthly spend by 40%. Exactly what was promised." — CTO

The outcome

The platform now spans two clouds, scales with demand instead of paying for the peak all month, and no longer has a single provider as a single point of failure. The migration completed without a minute of customer-facing downtime, and the monthly cloud bill fell by 40% once workloads were right-sized and automation removed the waste.

• 40% reduction in monthly cloud spend.
• Zero downtime throughout the migration.
• No single-provider point of failure — resilience across AWS and GCP.
• Automated CI/CD replacing manual, high-risk releases.
• Full observability, so cost and performance are now visible and tunable.