Database migration cost decisions.

Most database migrations are justified on a compute comparison and then quietly exceed it. The fix is to model the whole cost shape before you commit. Here are the decisions that determine whether the migration saves money or inflates the bill.

Managed or self managed: which is cheaper?

A managed database removes patching, backups, failover, and much of the operational toil, and that is worth a premium for most teams. But the premium is real and grows with scale. For a large, steady, well staffed workload, running the engine yourself on compute with reserved capacity can undercut the managed service, because you are no longer paying the management margin on every instance hour. For smaller, spiky, or thinly staffed workloads, the managed service almost always wins once you price the engineering time it saves and the outages it prevents. The honest comparison is total cost of ownership: managed price versus self managed compute plus storage plus the loaded cost of the people who would operate it. Decide per workload, not by policy.

BYOL or license included?

For commercial database engines, the licence is often the largest single cost, and how you carry it matters as much as where you run it. Bring your own licence reuses entitlements you already own, so you avoid paying for the licence a second time inside the hourly rate, which suits organisations with a large existing licence estate. License included bundles the licence into the price, which is simpler and suits teams without entitlements or with fluctuating needs. The Oracle case is the one to model carefully: on OCI, BYOL combined with Support Rewards can offset a portion of on premises Oracle support fees as you consume, and the choice between Base Database and Autonomous Database changes both the operational model and the cost. Run both scenarios with your actual licence position before assuming either is cheaper.

What storage and transfer costs get missed?

This is where compute only estimates fall apart. Database storage is billed on provisioned capacity and, increasingly, on provisioned IOPS and throughput, so a high performance tier chosen by default can cost more than the instance itself. Read replicas and multi zone deployments duplicate storage and generate cross zone traffic that bills per gigabyte. Cross region replicas add cross region transfer on top. Backups and snapshots accumulate on their own retention clock. And egress, moving data out to applications, analytics, or another cloud, is a recurring line that a one time migration estimate rarely captures. Size the storage tier to the workload's real IOPS, place replicas deliberately rather than by default, set backup retention to the actual recovery requirement, and forecast egress from real traffic patterns.

How do commitments apply to databases?

Steady database workloads are good candidates for commitment discounts, but the same risk adjusted discipline applies. Reserved capacity for managed databases on AWS, Azure Reservations for the database tier, GCP Committed Use Discounts, and OCI Universal Credits drawdown all reduce the rate on capacity you are confident will persist. Because a production database is rarely turned off, the stable baseline is usually large and well suited to commitment, but a migration in progress is exactly when the forecast is least certain. Commit the part you are sure of, leave the part still being sized on demand, and revisit coverage once the migrated workload settles. A commitment placed on a database you are about to rearchitect is the classic way to strand a discount.

Model the whole cost before you migrate

We model database migration economics end to end across AWS, Azure, GCP, and OCI, managed versus self managed, BYOL versus license included, storage, replicas, and egress, with zero provider commissions and no recommendation that risks production. Our guarantee: we reduce your cloud spend or we reimburse our service fee. Pricing is either a Fixed Fee scoped up front or Gainshare, a share of verified savings with no retainer and no risk.