AWS has done its quarterly housecleaning / “Googling” of its services, and deprecated what appears at first glance to be a startlingly long list. However, going through them put my mind at ease, and I’m hoping this post can do the same for you.
What Got the Axe
19 services are mothballed (“maintenance mode”), four are being sunset (“you can’t use these anymore after an upcoming date), and one is being end of supported (“it’s finally dead”).
A few are alarming: something like “Cloud Directory” seems like it’d be hard to replace, until you think about it and realize that you’ve never used it. Now that you really think about it, you don’t know anyone who has, either.
The ones that really jumped out to me are “Amazon Glacier,” “S3 Object Lambda,” “Snowball Edge,” and “CodeCatalyst.”
The Ones That Matter
Glacier is a red herring. Once upon a time Glacier was its own service, with its own APIs. Now, it’s an S3 storage class. What they’re doing is removing the ability to interact with Glacier via its own APIs, which frankly have always been profoundly annoying to work with.
S3 Object Lambdas have always been a bit weird. You can still have Lambdas operate on S3, and at least actual Lambdas are likely to see service improvements; Object Lambdas have been moribund for years.
CodeCatalyst was a big deal when it launched, and afterwards nary a peep was heard from it, either from customers or from AWS. This could have been something, but the will to make it that thing clearly has departed AWS along with some of its better talent.
That leaves Snowball Edge. This is a weird one, because a bunch of customers have run local EC2 instances on them, as well as using them for data transport jobs. Those customers can continue to do so (for now, at least), but if you’re architecting something new that leverages this I’d suggest making other plans.
Everything Else
A bunch of the modernization stuff that’s being Googled has simply been dragged into AWS Transform. New service marketing, same capabilities, and to top it off if you’re doing a migration you at least aspirationally like to think you won’t be doing it forever; finish your damned migrations already.
IoT Greengrass V1 let you run Lambdas on your own gear, and v2 has been out for many years. I do give this one a bit of a questioning side-eye, since it’ll require updating deployed things in the customer field, but… if it’s running detached entirely and hasn’t been updated in this long, keep on going, I guess?
Systems Manager Change Manager and Systems Manager Incident Manager are being wound down, with replacements ranging from “other Systems Manager capabilities with equally bad names” to “do what sensible people do and use a best in class third party option instead.”
The Bottom Line
Most of these deprecations appear to me to be the rotten fruit of the AWS “launch a new service to solve problem X” approach that persisted for far too long. It was clear that not all of these would be commercial successes, and I’m optimistic that clearing out their shambling corpses will let Amazon put more effort into the fewer things that actually matter for customers.