Microservices and multicloud strategy at Snap

I really enjoyed this article on migrating from a single region application monolith at Snap to a modern microservice architecture:

• How Snap rebuilt the infrastructure that now supports 347 million daily users

Thoughts:

• I like how Snap seems to be treating GCP and AWS as commodity compute resources with their approach. They’re picking the one that works most efficiently in their supply chain at a given time and adjusting as they go to reduce their end cost.
• Anyone who’s planning to take a similar approach to Snap should remember their learning lesson about trying to write applications on one layer and then choosing the cloud provider best suited to a workload. I’d like to read the whitepaper on why this was challenging and why it would have been so expensive to implement.
• Spinnaker sounds like it’s working out for Snapchat’s deployment use case. I’d like to understand what other CD solutions they evaluated and why then went with Spinnaker.
• FSN (Feelin-So-Nice) is an awesome internal project name. FSN sounds like something serious and technical, and no one outside the engineering team would probably ever question the name. Next time I need inspiration for an internal project name, perhaps I’ll pick the first letters of the most recent song I’ve listened to.

[Multi cloud image credit]