Leaving the Cloud Part 2

Key takeaways

Moving off the cloud can lead to significant cost savings and operational control, especially for medium-sized companies.
Proper planning and the use of simple, well-understood tools like Docker can streamline infrastructure transitions.
The myth that cloud services are cost-effective and require less operational effort is often overstated; in many cases, owning hardware can be more efficient and predictable.
It is possible and practical to manage your own data centers with support from experienced vendors, preserving valuable technical skills.
Organizations should critically evaluate their cloud costs and consider alternatives, including leasing hardware from vendors, to optimize expenses and control.
Understanding and rebuilding hardware management expertise can be a strategic move in a company's long-term infrastructure planning.

Summary

The Motivation Behind Leaving the Cloud

This section covers 37signals' decision to transition off cloud services, driven by issues like high costs, lack of independence, security concerns, and the realization that operational work in the cloud isn't significantly less than managing hardware independently. The company's initial plan to seek help with the transition failed due to mismatched service agreements, reinforcing their resolve to bring infrastructure in-house. The motivation is compounded by the mounting costs—they are spending around $38,000 weekly on cloud services—which prompted a reassessment of their infrastructure strategy.

The Transition Strategy and Tools Used

Here, the podcast discusses how 37signals opted for a simplified, more manageable approach using Docker and a custom tool called Kamal to facilitate their move out of the cloud. They started by moving low-criticality applications like Tadalist and Writeboard, which served as templates and testing grounds for the process. The company emphasizes that their infrastructure transition is prioritized based on the criticality of applications, starting with the least critical. They aim to be fully out of the cloud by the end of summer, with a clear plan for moving higher-criticality apps later.

Operational Changes and Cost Savings

This section highlights how moving off the cloud will streamline operations by reducing the time spent on managing cloud infrastructure, dealing with incidents, and handling ongoing upgrades and cost control in AWS. They project significant financial savings—around $7 million over five years—by owning their hardware, and stress that for companies in their size range, the cloud often does not provide the supposed operational benefits, especially at their scale. Additionally, they discuss the advantages of being familiar with managing their own data centers and the opportunity to rebuild in-house expertise.

Misconceptions About Cloud Costs and Expertise

The discussion debunks myths surrounding cloud computing, especially the idea that it is inherently cost-free or requires minimal operational effort. David notes that their cloud expenses are heavily optimized, yet the costs remain high and variable—particularly problematic for smaller companies. The company has aggressively minimized their AWS costs through reserved instances, enterprise agreements, and other leverage. They caution against becoming locked into proprietary cloud services and advocate for exploring alternative options such as leasing hardware from vendors like Hetzner, Digital Ocean, and Deft, making the case that owning hardware is more manageable and less costly than many assume.

The Broader Industry Perspective and Knowledge Preservation

This section reflects on how cloud has transformed the industry, noting that many companies have lost the skills and knowledge to run hardware independently due to the widespread migration to cloud services. David emphasizes that owning and managing hardware is not complex or outdated, especially with support from experienced vendors. They aim to demonstrate that alternatives to cloud are feasible and that organizations can retain or rebuild operational expertise by doing so. The discussion also touches on the availability of affordable, lease-based infrastructure options that can ease the transition and reduce costs for organizations.

Show notes

In October 2022, 37signals announced they would transition off cloud services in a post titled "Why We're Leaving the Cloud."

Despite the advantages of the cloud, the downsides were too significant for the company. In this follow-up episode to Leaving the Cloud on REWORK, co-founder David Heinemeier Hansson and Director of Operations Eron Nicholson discuss the progress 37signals has made since they decided to leave the cloud.

Listen in as they delve into the pros and cons of using the cloud versus owning your hardware and some of the misconceptions people have about cloud costs and expertise.

Tune in as the pair shares insights from their experience running web applications at scale and offer advice, tips, and tools for startups and businesses looking to optimize their infrastructure.

Join the discussion as they explore the importance of getting real about the costs of the cloud for small businesses.

Show Notes:

[00:49] - David reflects on the progress made since the announcement to transition off of cloud services. And why their initial plan for help with the transition didn't work for 37signals' needs.

[01:37] - How the experience with the failed deal to get help with the cloud transition reinforced the idea of why 37signals wanted to leave the cloud.

[02:33] - David shares some tools that can be used to move out of the cloud, including Docker and Kubernetes.

[04:21] - Why 37signals opted for a new tool they built called Kamal to assist in their transition and their timeline for being entirely out of the cloud.

[06:11] - Bleeding money: The $38,000 motivator to speed up the timeline and free up cash that could be better spent elsewhere in the business.

[08:33] - The tools: Eron and David discuss the tools (Docker and Kamal) they've used to successfully move two apps, Tadalist and Writeboard, from the cloud to their new system and the template in place for moving the rest of their apps.

[11:08] - David discusses the criticality ladder they have in place to prioritize their move off the cloud, beginning with the lowest criticality app (Tadalist) before moving on to higher criticality apps like Hey, which deals with email and must always be available to users.

[13:23] - The app transition plan for the next few months.

[13:58] - Eron explains the challenges of staffing up for a project of this magnitude. He also highlights how many of their current AWS-related processes will become streamlined once they have their own infrastructure.

[16:00] - David shares what the cloud is great for, why it doesn't work great for middle-range SaaS companies, and the main reasons for 37signals for bringing their infrastructure back in-house.

[19:31] - Eron explains their advantage in moving back to their own infrastructure and why he hopes that their experience will encourage others to consider moving away from the cloud as well.

[20:24] - Blowing the notion that the cloud is the only option to smithereens: David reflects on the lost knowledge of running your own hardware and the vendors who can help.

[23:17] - Eron shares how you can lease from vendors like Deft, Digital Ocean, or Hetzner until it makes sense for you to buy your own servers.

[23:46] - David discusses the importance of exploring alternatives to AWS, breaking the misconception that the cloud is the only option, and exploring the most cost-effective solution for your company.

[26:04] - David responds to a comment suggesting they could have saved money by optimizing cloud costs and the possibly "crippling" variability of cloud costs that is a major concern for smaller companies—a risk eliminated by owning your own hardw