TL;DR version: Think building a data center from scratch is simple and cheap? If you’re ever planning to scale, it might be time to think again.
Initially, we thought building and maintaining a cluster of servers from scratch would be the cost-effective and easy way to go—especially as we already had a well-operating NOC for other products. But that doesn’t account for DevOps and rapid growth. The second you incorporate virtualization, storage and database software, the costs start rising. Once we dove a little deeper, we realized that building high availability requires even a second colocation and more manual employee efforts for managing infrastructure and databases than we wanted to invest with our automation mindset.
If continuous delivery was going to be more than just a catchphrase, it was time to crunch the numbers and weigh our options. We’re performance hackers here—always eager for feedback so we can move to the next (see: better) iteration fast.