What is the difference between on-demand and reserved instances?
In the world of Oracle Cloud Infrastructure (OCI), the choice between On-Demand and Reserved instances isn't just about price—it’s about balancing financial risk against availability guarantees.
While other clouds (like AWS) bundle a discount and a hardware guarantee into one "Reserved Instance" (RI), Oracle splits them into two distinct concepts. Here is how they compare.
On-Demand is the "buy it when you need it" approach. It is the default for most users.
Commitment: Zero. You pay only for the seconds the server is running.
Availability: There is no guarantee that a specific server type will be available in a specific data center at the exact moment you want to click "Launch." In rare cases of high regional demand, you might see an "Out of Capacity" error.
Pricing: You pay the standard "List Price." This is the most expensive way to run a server per hour, but the cheapest if you only need it for a few days.
Best For: Development, testing, seasonal spikes, or new projects where you aren't sure how much power you'll need.
In OCI, "Reserved" refers specifically to Capacity Reservations. This is about making sure the "doors are open" when you arrive.
Commitment: You "claim" a specific slice of hardware (e.g., 20 AMD E4 cores) in a specific Availability Domain.
Availability: 100% Guaranteed. Oracle sets that hardware aside specifically for your tenancy. Even if the rest of the data center is full, your reserved slots are waiting for you.
Pricing: You pay for the capacity whether you are using it or not.
If the VM is running: You pay the standard rate.
If the VM is NOT running: You pay a discounted "Capacity Reservation" fee (usually 85% of the standard OCPU price).
Best For: Mission-critical apps, Disaster Recovery (DR) sites, and large-scale production environments that cannot afford a "capacity full" error.
If your goal is simply to get a discount (which is why most people look for "Reserved Instances" in other clouds), OCI uses Annual Universal Credits.
Instead of reserving a specific server, you commit to a specific dollar amount per year. In exchange, Oracle drops the price of every instance you run—on-demand or reserved—by a significant margin (often 20% to 35%).
| Feature | On-Demand | Capacity Reservations (Reserved) |
| Billing Starts | When the VM is "Running" | As soon as the reservation is created |
| Availability Guarantee | Best Effort | Guaranteed |
| Cost for Idle Server | $0 | ~85% of OCPU hourly rate |
| Flexibility | Change shapes anytime | Fixed to a specific shape/region |
| Termination | Stop/Delete anytime | Must delete the reservation to stop billing |
Use On-Demand if... you want the lowest total cost for workloads that don't run 24/7, or if you are comfortable with the very small risk that a specific shape might be temporarily unavailable.
Use Capacity Reservations if... you are running a "Production" environment or a "Disaster Recovery" site where the server must be able to power on instantly, no matter how busy the cloud is.