Thanks David,
Do you think the Free Tier one might be suitable ?
Although the sizing , a minimum of 2 GB is required.
Still trying to rationalise the cost and ability to put on EC2.
No, the free tier will run out after 1 year and not available to Windows instances. Even if it is, t2.micro will not fit your Windows instances since Windows are inherently have a higher compute consumption baseline. All micro instances are for bursty purposes and will not provide you with sufficient network, I/O and cpu performances.
You can't justify workload based on looking at specs alone. You must consider AWS from a TCO manner. For example, if you have a traditional on premise setup, how you factor in power consumption, and PSU, and the need for redundancies. How many the cost of having an engineer or 2 for network setup and configuration, storage and backup ? You also need to consider what happens to the server after a few years. I know depreciation works in 3-5 years, but still you need to get rid of the server, and would you consider having it reuse ? In 3-5 years, the hardware would have evolved a lot.
When performing sizing for traditional hardware, people normally like to spec higher than necessary due to the nature of future proving. You don't have to do it for AWS because you can always upgrade your instance by scaling vertically first (unless you have hit the ceiling).
Also instead of getting a powerful server than only operates at 20-30%, you can always make AWS instance work at 60-70% workload using a smaller instance. That's because you can always upgrade later at no cost.
However understanding AWS cloud usage and sizing is very important. You need to know what is your load with respect to the performance of the machine. This question is always a tricky question because no one can tell you exactly except through real usage.
With on premise solution, you can't test this out without first procuring from your vendor. Make the wrong choice and you have a over powered system that is sitting there rotting away. Underspec and you will cursed upon and looks bad on your KPI.
So the idea is first run an on-demand of your application on an AWS instance for 3-6 months and observed how the load is by monitoring your cloud watch. After that, scale up or scale down as accordingly to allow 60-70% usage. Once you are comfortable with the machine type, procure a long term of 1,3 years reserved instance of that type to enjoy cost saving.
If in any exceptional case you need to temporary size up for some intensive job, just restart the server with a different instance type and temporarily pay on-demand price for that hours you use, then revert back and AWS will charge you using reserved instance charges again.
When you have more and more different variety of systems in AWS, you will realise how easy to manage and reallocate resources it is versus on-premise setup.
