My thoughts on “Even Amazon can’t make sense of serverless or microservices”
I just came across an interesting critique of microservices and serverless architectures, based on a case study from Amazon’s Prime Video team.
Even Amazon can’t make sense of serverless or microservices
The author, David Heinemeier Hansson CTO and co-owner of Basecamp and Hey.com writes in his post that Amazon switched from a serverless, microservices architecture to a simpler monolithic system, and it resulted in a huge 90% reduction in operating costs. He uses (rightfully) this example to argue that microservices and serverless architectures might not be the best choice for everyone, as they can sometimes overcomplicate things.
But it’s essential in my opinion to remember that this is just one example, and it might not apply to all situations. It’s also important to note that the authors view of microservices and serverless architectures seems quite negative, while there are many successful cases of these technologies being used in various industries. Look at his latest writings https://world.hey.com/dhh
The article also talks about how Amazon originally used a service-oriented architecture (SOA) and how it evolved into microservices. While the post points out the benefits of SOA at large-scale companies, they also argue that microservices can be “toxic” and a “zombie architecture.” But, we should keep in mind that microservices can be really helpful when used correctly and in the right situations. I do agree it is not for everyone and people tend to do things in extremes.
There is also a quote from Kelsey Hightower, a big name behind Kubernetes, to further criticize microservices. Although Hightower’s quote highlights some potential issues with transitioning to microservices, it doesn’t mean that microservices are a bad idea altogether. His full article is here.
The article in the way I read it criticizes microservices and serverless architectures, suggesting that they can sometimes make systems more complicated than they need to be. While the Amazon case study shows some potential drawbacks, it’s important to remember that these technologies can be effective when used in the right context. The original post is in my opinion written in a provocative tone that might be a bit too general. That being said, I think I understand why the tone in the post is used, I can understand how it generates a reaction from both sides of the microservice debate.
What I would like to see is a future post that addresses how the new route Hey.com has chosen after architecting off AWS is going. And not just the cost savings.
It would be interesting if there are things from the public cloud and microservices approach they perhaps are missing or decided to keep. I don’t believe it’s a all or nothing discussion. I do agree that too many follow the herd and jump on serverless, x as a service and architect for a-platform-world-dominance when they could do as Hey.com is succeeding at, while saving a truckload of cash.
Just looking for some nuances. I do enjoy reading David Heinemeier Hanson’s posts and recommend them. https://world.hey.com/dhh
You can find more posts about people, process and technology on my blog https://fogplug.com/blog there is also a Podcast, https://fogplug.com
