-
Hello tyzh,
Thanks for sharing your thoughts.
I agree with you on the part where no spot instance is available means no added capacity. However, there’s a variety of workarounds that you can do to offset that effect for the sake of optimizing cost. First, you can diversify the instance types the application uses so there’s less chance of capacity running out. Second, as you said, you can mix on-demand and spot, with on-demand being the primary capacity so you can sustain capacity. Third, you can use ASG and enable instance hibernation so an instance resumes faster when disruption occurs. The scenario isn’t really that far-fetched from reality, as it’s a common case in the industry. Spot fleets are used more frequently in production environments than you might think.
Let me know if this helps.
Regards,
Carlo @ Tutorials Dojo
