About AWS Fargate

I'll address your questions about AWS Fargate and how it would handle your Trino workload:

1. Scaling with AWS Fargate: AWS Fargate doesn't directly handle scaling of your application. Instead, it works in conjunction with Amazon EKS (Elastic Kubernetes Service) to manage the scaling of your Trino workload. You would need to configure Kubernetes' Horizontal Pod Autoscaler (HPA) or Vertical Pod Autoscaler (VPA) to define how your Trino deployment should scale. These autoscalers would determine when to add or remove pods based on metrics like CPU utilization or custom metrics you define.

2. Resource allocation time: When your workload requires more memory (scaling from 4GB to 6GB), Fargate doesn't need to take any specific action. Unlike traditional EC2 instances, Fargate allows you to specify the exact CPU and memory requirements for each task. When you deploy your Trino workload, you would specify the maximum memory requirement (6GB in this case). Fargate would then allocate a task with those resources immediately, so there's no delay in providing the maximum resources you've specified.

3. Impact on workload during scaling: With Fargate, your workload doesn't stop or pause during scaling events. Since you've already specified the maximum resources (6GB of memory in this case), your application has access to these resources from the start. If you're using Kubernetes autoscalers to add more pods, new pods would be started without interrupting existing ones. The same principle applies to scaling down - existing pods continue to run until they're terminated by the autoscaler.

4. Pricing and resource allocation: Based on your requirements (2 core CPU and 4-6GB of memory), Fargate would use the 2 vCPU configuration. The closest match would be:

2 vCPU with 6 GB memory

In Fargate, 1 vCPU is equivalent to 1 CPU core. The vCPU allocation does matter for performance, as it determines the processing power available to your workload. If your Trino deployment is CPU-intensive, having 2 vCPUs should provide similar performance to your current 2-core EC2 setup.

Remember that with Fargate, you're billed for the resources you request, even if your application isn't using them fully at all times. So, you'd be paying for 2 vCPU and 6 GB of memory continuously, which allows your application to burst up to 6GB when needed without any scaling delay.

Sources
Serverless Compute Engine – AWS Fargate Features – Amazon Web Services
Community | Leveraging AWS Fargate for Containerized Workloads
Tutorial: Getting started with AWS Batch and Fargate orchestration - AWS Batch
AWS Fargate for Amazon ECS - Amazon Elastic Container Service