Compute
Runs general-purpose workloads — virtual machines, bare-metal servers, and container hosts that execute application code.
The most general primitive — most other primitives are specialised compute.
Description
Across clouds, Compute is offered as resizable virtual machines (AWS EC2, GCP Compute Engine, Azure Virtual Machines) backed by broadly similar instance families for general-purpose, compute-, memory-, and GPU-optimised workloads. The main differences are conceptual rather than functional: each provider has its own instance naming and sizing scheme, its own regional and availability-zone model, and its own pricing levers (on-demand, reserved or committed-use, and spot/pre-emptible discounts). On-premises Compute is instead a self-managed physical server (e.g. Dell, Lenovo), where you own capacity planning and hardware lifecycle rather than renting elastic capacity. To the best of our knowledge these are equivalent primitives; pick based on existing footprint, regional reach, and operational model rather than raw capability.
Capabilities
- Run arbitrary OS and application workloads
- Scale vertically (size) and horizontally (count)
- Attach block, file, and object storage
- Place behind a load balancer for high availability