
AWS compute: virtual machines for every workload, from microservices to in-memory databases — with Graviton processors for the best price/performance.
EC2 is the building block of the AWS cloud: hundreds of instance types (compute, memory, storage, GPU) in dozens of sizes, up in minutes and paid by use. With Graviton (AWS's custom ARM) performance per euro rises noticeably for compatible workloads.
M for general purpose, C for compute, R for memory, I for NVMe storage, G/P for GPUs: the choice affects costs and performance.
AWS's ARM processors (now Graviton4): same services, less spend — the first thing we evaluate in an assessment.
On-demand for the unpredictable, Savings Plans for the stable, Spot for the recomputable: the mix is where you really save.
AWS's hardware hypervisor: minimal overhead, security in the silicon and tenant isolation.
Instances are chosen by family (M general purpose, C compute, R memory, I storage, G/P GPU) and generation: the suffix matters — 'g' is Graviton (ARM), 'i' Intel, 'a' AMD. Underneath it all is Nitro: a minimal hypervisor on dedicated hardware, freeing near bare-metal performance and isolating security. Purchasing is a strategy: on-demand for the unpredictable, Savings Plans/Reserved for the base, Spot (up to -90%) for the interruptible.
The first step of migration: the VM as it was, but elastic and replicable across zones.
Capacity that switches on for the campaign or month-end and off afterwards: you pay for use, not for the peak.
Entire environments created and destroyed on command: testing without dedicated iron.