Spending in the Clouds

Basecamp has cut back its reliance on Amazon and Google, but there’s one area where it’s tough to find alternatives to Big Tech: cloud services. Even so, there are ways to cut spending on this $3 million annual expense while keeping the company’s apps running smoothly. In the latest episode of the Rework podcast, Blake Stoddard on Basecamp’s Ops team talks about how he volunteered to look for savings on cloud services and really delivered—to the tune of over a half-million dollars.

A transcript of this episode is also available. And if you like what you hear, be sure to subscribe to Rework in your favorite podcast app so you get all of our new episodes as soon as they’re released.

17 thoughts on “Spending in the Clouds

  1. I’m so confused. The core premise of this article is to be less reliant on Amazon & Google. The interview goes into depth about how they have reduced their Amazon bill by $700k.

    Then the interview ends with “we’re going to move Basecamp 2 and Basecamp 3 to the cloud at some point in the future … so our bill will go back up”.

    What !?!? You’re going to move MORE to Amazon / Google?

  2. >> “ Basecamp has cut back its reliance on Amazon and Google, but there’s one area where it’s tough to find alternatives to Big Tech”

    There’s plenty of alternatives. Just to make a few, instead of using Amazon it Google – you could use:

    Digital Ocean

    All of which aren’t surveillance companies like Amazon/Google and they are radically cheaper as well.

    1. I moved my company off AWS ($1.6M/years spend) over to Vultr.

      Our spend now is only $480k/year (70% cost savings) and our service is more performant.

      Note: I’m not affiliated in any way to any cloud provider.

    2. Hey Frank!

      When we started this migration three years ago, we were going after a few things: a) less reliance on managing our own hardware, and b) less reliance on managing the software that our apps run on (MySQL, Redis, etc.) so that as we continue to work on new projects, we don’t have to grow the Ops team to keep up. The legacy apps that I talk about in this episode are running on AWS ECS and were the first things we moved to any cloud provider.

      As part of the “lets-skate-to-where-the-puck-is-going” mentality, we also wanted to containerize those apps so that we could get better binpacking, etc., but running your app in a container isn’t helpful unless you have some orchestrator to actually run it. At the time, the only options on that front that were commercially viable and didn’t require us to deal with provisioning and running Kubernetes (or an alternative) ourselves were Google’s Container Engine (now GKE) or AWS ECS. Similarly with managed MySQL (and even object storage to an extent), the smaller providers just didn’t have performant solutions — DO didn’t launch managed MySQL until this year, and then there’s a question of network connectivity between cloud providers and our data centers and the smaller providers simply don’t have the DirectConnect-esque options that we needed.

    3. I do believe that AWS is the market leader in the cloud industry and I do believe that it stays this way. As, they are more into the innovations and bringing new things to light.
      However, the other hosts you mentioned has done a pretty well job for them in past years. Especially Vultr, Linode and DigitalOcean. I have personally used the managed instance of all of them as their managed instance is powered by Cloudways and I have found it quite an interesting option. Just check the link below:

  3. The Ops team can speak in more detail to this if they choose, but it’s my understanding that they’ve looked into/are looking into many of the alternatives listed above. Switching away from Amazon and Google is not as simple for cloud services as it is for other areas where we’ve given them our business.

    And to Steve’s point above, perhaps when the company is ready to move BC2 and BC3 to the cloud, the Ops team will have found an alternative provider that they feel really good about. And then the overall cost of cloud services will go up, but not necessarily with Amazon or Google.

    The episode fits into a larger topic we’ve been addressing on the show and elsewhere, which is reducing our overall reliance on Big Tech and standing for something different. It’s not possible to quit cold turkey, which is why we still spend money with the Amazons and Googles of the world—although we’re trying to do less of that. We also think it’s worthwhile to document the process as it unfolds, even though it doesn’t make for the tidiest or most airtight story.

  4. What caused Basecamp to even move to the cloud?

    It was mentioned in the podcast that the first 12+ years of Basecamp existence was self hosted and only 3 years moved to the cloud (and moved from cloud to cloud AWS -> Google -> back to AWS)

    1. Hey Ted!

      Combination of a few reasons, partly a “skate to where the puck is going” mentality (like I mentioned in another comment), partly wanting to be able to scale better without being tied to bare metal, partly to get away from managing our own hardware and certain software pieces so that we can continue to grow the number of projects we work on without having to grow the Ops team.

  5. Hi there, it’s the first episode I listen from Rework. I find Blake’s voice sounds like a late 90’s 16Kbps mp3, I’m glad there’s a transcript to help. I’m sure you can do better to carry the stories of your guest. Thanks!

    1. Hey! Sorry about that, my mic misbehaved and I had to fallback to one on my laptop which clearly is no studio quality mic. If I’m on Rework again, we’ll get it fixed up!

  6. I too would love to read more about why Basecamp even went cloud, given that hadn’t for such a long time and dedicated is so much cheaper.

  7. I am honestly very surprised by the numbers in this story.

    $2 – $3 million already seemed crazy, and then Blake said that Basecamp 2 and Basecamp 3 — presumably a very large chunk of computing needs — are not hosted on cloud services yet.

    I don’t have cost-saving suggestions or stories of moving to some other service and saving millions. My assumptions about how much a company like Basecamp would need to spend on hosting were simply wrong, by an order of magnitude or more.

    1. Hi, according to their website, there are 3 millions accounts signed up. So that’s 1 dollar per account spent in the cloud. An account is 1188$/year. Not that bad I find :).

  8. Now I regret I did not apply to your director of operations position. I am so used to having a super tight budget I would have made a killing on not cost-optimized system 🙂

Comments are closed.