Fairly high-level, but of interest. The five lessons are:
- Understand and embrace the differences between a cloud environment and a normal server environment.
- Build your systems to accomodate failure at every level.
- Introduce planned, random failures to test your redundancy (make failure of individual components a daily occurrence)
- When researching new technologies, test them at full scale.
- Be completely committed to the process.
Netflix were in the fairly interesting position of already having a huge service running in a "normal" datacenter before deciding to migrate to AWS, so this advice may not apply to everyone, but no doubt there will be many other companies looking to migrate large, existing services to a cloud infrastructure.
