As some organization expand their cloud footprints, they realize that renting infrastructure with no strings attached can rack up substantial bills over the long haul.
But for businesses that run predictable workloads, long-term commitments are a good way to cut cloud computing costs. There are two popular options to choose from: Amazon Web Services (AWS) Reserved Instances and Google Compute Engine (GCE) committed use discounts. Both options provide a cut to on-demand prices in return for a one- or three-year term commitment from users. However, there are significant differences.
AWS Reserved Instances vs. GCE committed use discounts
AWS offers two kinds of Reserved Instances: Standard and Convertible. The Standard option lets enterprises adjust instance size, but not the type, while the Convertible option allows users to exchange one instance type for another.
Rather than apply to an instance type, GCE committed use discounts apply to the total amount of CPU and RAM a user purchases. Users can mix and match instance and machine types and sizes. The discount also applies to custom machine types, which enable users to tailor the mix of virtual CPU (vCPU) and memory that they want.
Unlike AWS, which offers the greatest discount to Reserved Instances users who pay for the entire term upfront, Google just deducts from users' monthly cloud bill.
GCE plan details and pricing
GCE committed use discounts apply to any mix of vCPUs and memory that users purchase within a region. As with custom machine types, the aggregate memory must be within a range of 0.9 GB to 6.5 GB per vCPU. Discounts apply first to any custom machine types and then to a random selection of vCPUs from predefined machine types.
Google offers this example: Say an organization assigns 10 vCPUs and 30 GB to custom machine types and also uses two n1-standard-4 (4 vCPUs, 15 GB each) instances. If the organization commits to purchase 15 vCPUs and 13.5 GB of memory, the discount would apply to all 10 of the custom machine vCPUs and 13.5 GB of the memory. The remaining 5 vCPU discount would apply to five of the eight CPUs in the two standard VM instances.
Resources in a committed use pool don't qualify for Google's sustained use discounts. However, other resources in the same region not covered by commitments -- such as three of the standard vCPUs in the example above -- can receive lower rates for usage over 25% in a month.
Exact GCE committed use discounts vary based on whether admins make a one- or three-year commitment, according to Google's pricing documentation. Discounts max out at 57%.
Although the list price discounts are compelling, an organization's overall cloud computing costs will be dependent on the workloads and usage patterns for the applications it deploys.
Usage scenarios and recommendations
Usage-based commitments, in general, require predictable levels of baseline VM workloads, as well as knowledge of vCPU and memory requirements over time. Once you commit, you pay for the total usage, whether you need it or not. That said, compared to AWS Reserved Instances, GCE committed use discounts don't require organizations to micromanage instance selection and predict the VM characteristics they'll require three years into the future.
Usage-based commitments are unwise for sporadic or seasonal workloads, such as test/dev, quarterly or monthly batch processes and machine learning model training. Don't commit to a set amount of compute capacity when you don't know how much you'll need in the future or for how long.
Review usage-based discounts from AWS, Azure and Google
Optimize costs through cloud regions and availability zones
Evaluate Google's range of cloud instance types