Amazon Elastic Container Service (Amazon ECS) now supports CPU limits of up to 192 vCPU for ECS tasks deployed on Amazon Elastic Compute Cloud (Amazon EC2) instances, an increase from the previous 10 vCPU limit. This enhancement allows customers to more effectively manage resource allocation on larger Amazon EC2 instances.
Amazon ECS customers can define soft and hard limits for CPU and memory resources at the container level, and hard limits at the task level. Soft limits reserve resources on an Amazon EC2 instance for a container, while hard limits enforce maximum usage. For CPU specifically, the container-level hard limit acts as a ceiling and helps prevent resource contention when multiple containers are competing for resources using Linux CpuShares. The task-level CPU limit acts both as the reservation for the task and prevents any single task from consuming excessive resources during contentions. Customers can now specify up to 192 vCPU as the CPU limit for an ECS task, increased from the previous 10 vCPU, enabling more effective resource sharing across multiple tasks on larger sized EC2 instances. For example, on a c7i.48xl instance with 192 vCPUs, defining a 32 vCPU limit per ECS task allows running up to 6 tasks without resource contention from noisy neighbors.
You can use AWS management console, SDK, CLI, CloudFormation, or CDK to define the CPU limit for your Amazon ECS task definition. The new limit is now effective in all regions. To learn more, see documentation.
Source:: Amazon AWS