AWS Pricing — Where can you go wrong as a Newbie?

Amazon Web Services is overwhelming when you are new to the platform. It is a hard time finding the right offering for your problem in the ocean of solutions offered by the platform. Most of the times, when you are busy finding that right solution, you settle on an exciting offering that is easy to implement in your world and you completely miss the price tag attached to it till you are served the bill or you check the bill and see the spike because of your last purchase. Sounds like giving your credit card to your kids to shop and getting a shock once you get the credit card bill.

There are a few rules of thumb that needs to be checked as a starter :

1. Remember that any customizable solution that calls for synchronous jobs will have endpoints that can be called for real time analysis will always have more cost than asynchronous tasks
2. Some solutions have a per day price associated with their endpoints irrespective of whether they are in use or not. Turn off/ deactivate or delete these endpoints when you are not using it.
3. There are a number of solutions that can work for a given problem statement. One should spend energy in going word by word on the pricing of every solution and then evaluating them based on one time price and on going price. Some solutions will have one time higher cost but they would be very less to maintain in the long run. While some would provide free setup at the beginning, but you will end up paying more in the long run.
4. There are endpoint deployments that will have a fixed cost per day irrespective of usage. Most of these endpoints are associated with AWS API services that provide in built models to carry out specific functionalities

There are also set of good practices that can help you save in the long run. I am citing a few that I feel are easy to do and helps you save :

1. Delete end points when they are not in use. For example, if you have created an endpoint on an API service or SageMaker that is being tested for real time usage of the application. Do not forget to delete it when it is not being used.
2. Shutdown instances when you are not using them. For example, if using the AWS Sagemaker to build a model. While calling it a day, the best practice to stop making them charge you is to shutdown the Jupyter Notebook instance. This will prevent charges being levied even when you are not working on the model.
3. Logout of AWS before shutting down your system.
4. You can create a notification request in settings for lower slabs so that you are notified if you are being billed for an error from your side.

AWS is very kind in providing free learning hours and deployment to starters. If you follow the above, you can make use of these hours to create a great learning experience for yourself. I will keep updating saving tips as I understand the platform more.