Everyone went crazy over the service deprecations AWS silently announced over the past few weeks. There has been a lot of criticism and many engineers got serious trust issues as AWS never really used to discontinue services before.
I don’t understand this overly negative sentiment, I believe it was a smart management decision and will be a blessing for both AWS and all customers (except the very few that now need to migrate their applications — hey btw if you need any help with that, I’m here).
All of the deprecated services were barely used and poorly built compared to competitor products or the usual “AWS quality standard”. CodeCommit didn’t stand a chance against GitHub and GitLab. I think they knew CodeCommit wasn’t it for quite some time — its last update was this — 4 years ago:
I myself have never even heard about some of the other services. Here is a list:
- CodeCommit
- Cloud9
- S3 Select
- QLDB
- Cloudsearch
- SimpleDB
- Forecast
- Data Pipeline
- Snowmobile
All of them didn’t gain a lot of user traction, some didn’t get any at all. After AWS introduced the Snowmobile at re:Invent in 2016, I doubt that many saw it live & in action, apparently customers didn’t care about cool trucks and their exabyte-scale capacity.
There have been comments saying that all of these deprecations demonstrate AWS’ failure to execute, I believe it shows the opposite. AWS doesn’t need to be the market leader in every single product category; it’s okay to experiment, and it’s okay to fail. Well, as long as AWS doesn’t kill us-east-1, I feel like half of the entire internet is running on it.
AWS has never been famous for their developer experience (I still have my angry moments from time to time) — with CodeCommit and buying Cloud9 they tried to get there, but were quickly humbled by the fact that IaaS is at the heart of AWS — and not every-single-thing-that-has-to-do-with-cloud.
And that’s fine. Where I do think the criticism is justified is in how AWS communicated these deprecations. They silently announced they will discontinue a few services with generic messages that sound innocent and unobtrusive. I would have expected a direct and honest “We are killing some products that didn’t make us any money because — quite frankly — we’re not good at everything”.
Maybe killing these products wasn’t just a financial decision but also part of the agile-fetish every company started to adopt. AWS’ own interpretation of agile became a management principle called the 2-pizza-rule. By keeping teams at a small size of 6–8 (I am still struggling to understand how 2 pizzas can feed 8 people — and Americans for sure eat more calories than me), AWS hoped to increase efficiency and reduce time-to-market. In theory, this all makes sense — we all hate sitting in meetings with 20 people talking — and I’m sure this is one reason for AWS’ continuing success and growth over the past years. A side effect, however, is that small teams with a motivated PM could launch whatever niche product they could dream of — as long as they could make a good case for it and convince upper management about it’s scalability and massive potential.
So Agile caused a lot of useless products. The AWS product catalog (and their respective staff) grew massively, which increased operational overhead and maintenance cost, which is not very agile-like. It sounds too much like corporate, the biggest turnoff vocabulary for big tech — AWS still thinks they are a startup. Keeping the status quo was not an option. So, the products and people initiated by Agile were now let go — by Agile.
There is a coolness battle going on in big tech about who can lay off the most people, and AWS doesn’t want to lose this one — so keeping an endless amount of products alive was not an option…
There are also a lot of rumors around which service is dying next. For those worrying that services they depend on might be next on the chopping board, I believe this was rather a one-time cleanup rather than a new norm for AWS. AWS still isn’t Google, and it will never be.
So, calm your balls everyone, it’s gonna be fine.