The hidden costs of cloud

At Server Density, we just completed a 9 month project to migrate all our workloads from Softlayer to Google Cloud Platform. This started with a single service using Cloud Bigtable for large scale storage of time series monitoring data and culminated with our entire product now running on GCP.

The move went smoothly and we’re very happy to be on Google Cloud. Although they have fewer products compared to the market leader (AWS) we think that Google Cloud products are better designed, the pricing models are easier to work with and there are regular releases of innovative new features e.g. the recent Andromeda networking improvements.

That said, now the migration has completed we are reviewing our costs to ensure that we have a well optimised deployment. This is about finding savings that make a difference to the monthly bill as opposed to a few 10’s of $ here and there. Doing so has revealed some interesting “features” of how cloud infrastructure is billed.

Cost can be difficult to predict

This should be obvious but it can catch you out if you are coming from a legacy hosting provider where you contract for specific servers and services, thereby having a fixed bill each month. This contrasts with the pay as you go model of cloud whereby you’re paying for actual usage.

This approach initially appears more efficient because you are only incurring fees when the services are consumed but it necessarily leaves you open to unexpected cost spikes if…