AWS Placement Groups - how might two instances sharing the same hardware pose a problem?

Question

I have been thoroughly reading the AWS docs on Placement Groups - Amazon Elastic Compute Cloud. In the Partition as well as the Spread group sections, the use cases described seem to focus on failures that occur from having instances share the same hardware, for example:

Launching instances in a spread placement group reduces reduces the risk of simultaneous failures that might occur when instances share the same underlying hardware

1. I'm trying to wrap my head around how a failure in an application on Instance A would affect Instance B because of their shared hardware
2. On a similar note, if this side effect was possible, wouldn't other instances be affected as well? (that aren't mine).
3. What are some real-life situations that this might happen?

Answer 1

You've reversed the cause and effect.

The issue is not that a problem with the application on virtual machine A could cause a problem with the application on virtual machine B because the two VMs share hardware.

The issue is that if the shared hardware itself fails, then both A and B will potentially fail simultaneously because of the failed shared hardware -- which they both depend on.

Spread placement prevents two or more critical services from being -- coincidentally and unknown to you -- deployed on the same physical hardware, which is a situation that would increase the likelihood of simultaneous failures from a single hardware cause.

Hardware failures in EC2 are relatively rare, but hardware does occasionally fail or become degraded.