Autoscaling of EC2 Instances

Question

I am completely new for AWS and learning some services provided by AWS. At many places it is mentioned that EC2 comes with various flavor where one can decide whether instance they need should have more memory, CPU etc and accordingly can plan. On some other places even it is mentioned that EC2 auto-scale which means if load goes high then it brings up more instances and when load is less remove those instances. Now both looks bit contradicting to me as one is asking to plan resources whereas other is saying whether it's done or not auto-scaling will take care if your planned resources somehow will not be able to take the load. Help me to understand how auto-scale works and do i need to provision the instances till where auto scaling can add instances.

Answer 1

Most websites have variable loads during the day. If you're Netflix, most people watch in the evening. If you're Facebook, most people post at lunch. If you're Uber, most people want a ride at night.

In the traditional "data center" world, you'd need to buy enough hardware to handle the peak load and the rest of the day the hardware is not doing much. With the Cloud, however, you can auto-scale when the load is the highest and downsize during the quiet times. Plus, you only pay for resources when you use them, so you can save money when you don't need as many computers.

Therefore, the decision of which instance types will be based on your type of application (whether you need lots of RAM for caching or lots of CPU for number-crunching) and the decision of how many instances will be based upon how busy your system is at various points in the day.