# aws


09/29/2022, 12:31 PM
Hi, I’ve started testing deploying services in ECS/Fargate using AWS Classic with Python. It’s going well so far! - Am I correct to think of AWS Crosswalk as a simplified high-level api that you’re likely to grow out quite soon if you’re doing anything relatively complex? - I tried out the AWS Native provider at first, thinking it made sense to use it in a greenfield project. Every ‘up’ operation took at least one minute. (I may have made some silly mistake.) I switched to AWS Classic and it was fast. Will the AWS Classic provider be supported for a relatively long time going forward?


09/29/2022, 1:56 PM
• Crosswalk should support your needs, regardless of your use case. It's designed to encapsulate things for the most common use cases (stuff you'd nearly always want, like a VPC across 3 AZs), but also to be production-ready and flexible. Would be interested to hear more about your use case. • AWS Classic will be supported for quite some time to come. • Can you provide an example of a simple-ish AWS Native use case that was taking a long time? It's my understanding that AWS Native is more performant than Classic.


09/29/2022, 3:32 PM
Thanks for your reply Josh, that’s good to hear about the future of AWS Classic. I’ll get back to you about the other details tomorrow.
👍 1
AWS Native being slower than AWS Classic: This consistently takes around a minute to run ‘pulumi up’ on using e.g. eu-north-1:
Copy code
from pulumi_aws_native import s3

And this takes about 5 seconds:
Copy code
from pulumi_aws import s3

For this experiment I used the Pulumi service as the backend.
Ah, found: I did some other instantiation timing tests using AWS Native/Cloud Control: ec2.VPC: 20 seconds ec2.Subnet: 10 seconds ecr.Repository: 4 seconds So I guess there is a wide performance variation when using the AWS Cloud Control API. These consistently take a few seconds using AWS Classic.