SYSTEM Cited by 4 sources

Amazon EC2 (Elastic Compute Cloud)

Definition

Amazon EC2 (launched beta 2006) is AWS's virtual-machine compute service. Its runtime experience is tightly coupled to systems/aws-ebs for system disks and to the networking/storage offload stack (systems/nitro, systems/srd) for data-plane performance.

Seen in

• sources/2026-05-19-aws-how-synthesia-optimizes-generative-ai-video-inference-on-amazon-ec2-g7e-instances — EC2 chosen by companies/synthesia "because of the flexibility and control over the underlying hardware that the service provides" for its in-house latent-diffusion video generation models. First wiki appearance of the G7e instance family — NVIDIA RTX PRO 6000 Blackwell GPUs with 96 GB of GPU memory — positioned as a "cost-efficient option to serve GPU-memory intensive generative AI video models". Customer-controlled hardware is load-bearing for the optimisation: dual CUDA streams + pinned host buffers + double buffering require fine-grained CUDA-level control that managed inference services don't expose.
• sources/2024-08-22-allthingsdistributed-continuous-reinvention-block-storage-at-aws — EC2 ran on systems/xen until late 2017; EBS perf work is framed as improving the EC2 experience end-to-end.
• sources/2024-11-15-allthingsdistributed-aws-lambda-prfaq-after-10-years — Lambda launched on single-tenant EC2 instances; later moved to Firecracker micro-VMs packed on bare-metal Nitro hosts.
• sources/2024-07-29-aws-amazons-exabyte-scale-migration-from-apache-spark-to-ray-on-ec2 — EC2 as the raw substrate for Amazon Retail BDT's systems/ray compactor (Q1 2024: >10k vCPU-years consumed, clusters up to 26,846 vCPUs / 210 TiB RAM); forcing function for patterns/heterogeneous-cluster-provisioning — discover an instance-type set that meets the workload's resource shape, provision whichever are most available across AZs, application agnostic to CPU arch / disk / hardware specifics.