Compute
Compute Instance
A single virtual server you rent from a cloud provider, defined by its vCPU count, memory, and disk.
An instance is one virtual server: a bundle of vCPUs, memory, and disk that the provider spins up for you on shared hardware. AWS calls them EC2 instances, Google Cloud calls them Compute Engine VMs, and DigitalOcean calls them Droplets. Same idea, different branding.
Instances come in families tuned for different jobs. General-purpose balances CPU and RAM. Compute-optimized weights toward CPU for heavy processing. Memory-optimized favors RAM for databases and caches. Picking the wrong family is the most common way people overpay, either renting memory they never touch or starving a database of it.
The count and size of your instances is where most of a cloud bill comes from, so this is worth getting right. Start smaller than you think and scale up once you've measured real load, rather than sizing for an imagined peak.
Go deeper
- How to Tell When You Actually Need to ScaleThe metrics that mean something versus the ones that just look alarming, and what to do about each.