AWS Batch + MonteCarlo Workshop - Updates

[ruecarlo]

The workshop need an update in the following areas:

• Change the module that deploys the montecarlo notebook into a sagemaker using cfn
• Change the web-app deployment from the instance to an ECS of Fargate (or similar) on the same VPC where the instance would be created.
• Instead of EC2 Spot Fleet, use ASG with mixed instances and weights (now that ASG support weights).
• Implement a multi-process queue in the python code to span as many workers as processors/vcpus
• Benefit from Scale-in Protection in ASG with SQS workers; Adapt the code and workshop to show the benefits of the scale in protection and handling of visibility timeouts in SQS
• On the AWS Batch section, change to use Spot cost optimized fleet, and help with the exercise to instead of the adoption of SQS for submission, adapt the application to write into S3 and call from S3 on to AWS Batch (to demonstrate the most usual intake of batch workloads) (or at least make this an exercise)
• The workshop should be about Spot + AWS Batch best practices, although the background is on FSI and MonteCarlo the Montecarlo part should be Optional so we can re-use the same workshop in many contexts (i.e: EDA, etc) general async workloads, etc