CASE STUDY 02 / CLOUD & DEVOPS · SOFTWARE ENGINEERING · SECOPS
Lower cloud costs.
More frequent releases.
Cloud and delivery modernization for a software business facing rising AWS costs.
APPROXIMATE ANNUAL CLOUD SPEND
Before GCP
$1,000,000
After migration
$750,000
After optimization
$580,000
Interview estimates. Migration: ≈25% reduction. Further optimization: ≈$170,000 annually. Not normalized to a constant workload.
ENGAGEMENT FOCUS
AWS → GCP migration
GitOps & platform optimization
CI/CD & code quality
Continuing security operations
THE OBJECTIVE
Move a live platform safely, reduce cloud spend, and give the product team a more frequent release workflow.
THE CHALLENGE
Cost pressure exposed a platform problem.
Annual AWS spend had risen from roughly $750,000 toward $1 million despite resource analysis and optimization. The business needed a less expensive platform without interrupting service. Releases were happening roughly once a week.
An earlier move from Heroku to AWS had established GitOps. That captured configuration became the foundation for planning the next transition—not a reason to rewrite the application portfolio.
Use a known platform to move safely.
Reuse the GitOps foundation
Configuration already captured in GitOps made the GCP redeployment easier to reason about. Planning to cutover took approximately six months.
Prepare the cutover
Prepared database migrations repeatedly pulled deltas ahead of a coordinated midnight transition after peak traffic. The team reported zero downtime.
Optimize beyond migration
After the move, node, service, autoscaling, bandwidth, and network optimization reduced spend further. The cloud switch alone did not account for the full reduction.
SOFTWARE ENGINEERING
Make frequent releases part of the workflow.
CI/CD, build linting, and code-quality checks introduced practices that had previously been absent. These delivery improvements changed how the product organization released software alongside the infrastructure transition.
The estimated cadence moved from about one deployment per week to around 30 per day. This was delivery-workflow engineering—not a claim that changing cloud providers alone made releases faster.
CONTINUING PLATFORM WORK
Beyond the cloud bill.
In the continuing relationship, GitOps helped the team understand installed software and cluster state during vendor transitions. FullStory moved to OpenReplay; New Relic to Loki, Tempo, and Mimir; and Harness, formerly Split, to FeatBit. The engineering organization was around 100 people in this later phase.
The FullStory transition also had zero reported downtime. Recalled vendor contract reductions totaled roughly $257,000 annually before replacement hosting and operating costs; this is not audited net savings and is separate from the cloud reduction.
SECOPS
Keep readiness and remediation moving.
Security tooling transitions included Lacework to Falco, Prowler, and Wazuh, and Snyk to SonarQube and GitHub dependency scanning. These are implementation changes, not claims of equivalent security coverage.
Vanta setup, integrations, assessment, documented findings, and mitigation work supported SOC 2 Type I readiness. Independent attestation and report completion are not established. HolmesGPT was also used in canary version-upgrade workflows to support updates and security patching, separate from readiness work.
Lower spend. A different release rhythm.
≈42% lower spend
Approximately $1 million → $580,000 annually across migration and subsequent optimization.
0 reported downtime
A coordinated AWS-to-GCP cutover preserved service, according to the project interview.
≈30 deployments / day
Up from roughly one per week after CI/CD, build linting, and code-quality improvements.
Source: project interview. Costs and release cadence are estimates; 42% is calculated from approximate annual spend. Zero downtime is the reported cutover result, not an availability SLA. Vendor reductions are separate.
MORE CASE STUDIES
Moving to Azure without downtime →