What was being solved.
A digital agency needed a Magento environment prepared for a major client launch while keeping application and infrastructure responsibilities explicit. The launch date was fixed, the client relationship belonged to the agency, and the store could not be rehearsed against production traffic, so the preparation had to make the failure modes explicit instead of hoping the day would go well. The public record does not name the agency or disclose supporting traffic measurements.
How the engagement was structured.
Dependency and transaction mapping
Mapped web, application, database, cache, search, payment and deployment dependencies as one transaction path, from catalogue browsing through checkout to order processing. The uncached paths (cart, checkout and order writes) were treated as the sizing constraint, because cached catalogue traffic tells you nothing about what breaks first.
Capacity and caching path
Prepared a multi-tier architecture and caching plan against the expected demand pattern, with each Magento component (PHP workers, database, search service, cache and sessions) sized together rather than one layer at a time. Assumptions that could not be verified before launch were written down as watch items with an owner.
Deployment and rollback
Separated staging, production change, health checks and rollback decisions before launch. A change freeze, the approval route and the rollback conditions were agreed while everyone was calm, so no decision had to be invented during the event.
Operating ownership
Kept application-code ownership with the agency while defining infrastructure monitoring, severity rules and escalation for the launch window. The agency stayed in front of its client; the infrastructure path had a named engineer behind it.
The questions this work answers for your environment.
Size the uncached path
Cart, checkout and order writes bypass the cache. If those are not sized, the cache only decides how good the failure looks.
Size components together
Magento fails at the weakest of workers, database, search, cache and storage. One oversized layer does not rescue an undersized one.
Decide before the event
Change freeze, approval authority and rollback conditions are launch-day decisions that should never be made on launch day.
Split the ownership
The agency keeps the client and the code; the infrastructure gets a named owner and an escalation route. Both are written down.
What this page can claim.
The current public record supports an anonymised launch-preparation narrative. It does not provide an approved traffic multiplier, named client or independent uptime record, so this page avoids those claims.
A similar engagement starts with an environment review of the store and its launch constraints. Platform-specific sizing is part of the Magento hosting scope. See Magento hosting