For years, companies have been outsourcing their infrastructure to external service providers. However, 37signals, the developers of Basecamp and Hey, have recently decided to bring their IT back in-house. Farah Schüller, Senior Site Reliability Engineer at 37signals, spoke about the company’s practical experience and what other companies can learn from it.
Schüller explained that the decision was primarily an economic one. The costs of the cloud were becoming more significant for them, and in many ways were no longer in proportion to their use. Most of their applications had predictable traffic, number of users, and complexity, and the additional costs for any flexibility and abstraction were often no longer justified for their use cases.
The second point was that the operational effort increased relatively quickly. Automation, monitoring, CI/CD were built around the concepts and interfaces of the cloud service provider, which often changed or were completely replaced by something new. In the end, they also asked themselves a philosophical question: do they really want to place their operational business in the hands of an external service provider so much?
Schüller states that the key is for companies to draw on existing knowledge and adapt it to new challenges. As soon as possible in the development cycle of an application or service, it can be helpful to take stock: Does the application require a complex, dynamic infrastructure? Is user growth stable or predictable? Is it possible to have a reliable overview of the long-term costs and dependencies of outsourcing to the cloud?
Large providers can, but do not have to, do it better, and surprisingly, it is a highly individual consideration. It can be worthwhile not to be guided too much by trends and supposed simplicity. Companies can still tackle the path out of the cloud as long as they critically evaluate and ask the right questions.