Live data from Hacker News

AWS Is Asleep at the Lambda Wheel

lastweekinaws.com

31–36 of 36 posts

Re: AWS Is Asleep at the Lambda Wheel

#31
post #27

Earlier quoted context omitted.

They’ve provided custom runtimes which removes the need completely for managed runtimes giving you whatever version you like. They also did it in a way with Docker that’s simple and powerful. If I was an AWS exec I’d say managed runtimes are legacy, with the introduction of containers move to containers for new deploys. I’ve been critical of Lambda in the past and critical when people use Kubernetes/Docker because it…

> If I was an AWS exec I’d say managed runtimes are legacy, with the introduction of containers move to containers for new deploys. If they do that they remove one of the key benefits of AWS Lambda and that's that customers don't have to care about the underlying operating system and just have to care about their function code. And I'm sure if they'd would look at the numbers of how many AWS Lambda Functions use zip…

Of course 90% will use zip packages, container support is relatively new in the lifetime of Lambda.

Deprecating zip packages doesn’t mean zip packages no longer work, it means pretty much what they’ve done, ignored the managed runtimes with the view if the current runtime doesn’t do what you want, make your own with exactly what you need using tools you are likely already using with no vendor lock-in.

For things like Python most people don’t care about the underlying operating system from the projects in multiple orgs I’ve worked in. They go to Dockerhub, type Python, pick the image with the highest downloads and run pip install inside it. If you use Go, Rust, anything that creates a static executable you use “FROM scratch” with the only thing inside the container being the executable.

Those who do care about the underlying image are likely large orgs with a set of already blessed base images or people using things like Nix to build containers.

From a large org perspective and developer experience getting a “registry” for free because packaging as a standards compliant OCI image means instead of effectively creating your own registry on top of s3, you get a far more robust one with tagging, asset hashes, layering etc out the box. I never had a issue with zip archives but containers removes a bunch of roll your own boilerplate registry.

Re: AWS Is Asleep at the Lambda Wheel

#32

Seems like a pretty strong case against “serverless.” There’s always a server. You just have zero control over this one. So you’re stuck defactoring code to run an older version of Python because you can’t upgrade Python.

With managed runtimes it used to be you had no control. It was an opaque black box with complete vendor lock-in.

Now it supports Containers it’s a place you can run any compute you can think of as long as it finishes in 15 minutes, uses less than 10gb of memory and you want to trigger it using http, direct lambda invocation or from one of the various aws event sources such as s3, sqs, sns.

I used to avoid Lambda as it was an opaque black box. Now it’s a generic compute platform for short lived jobs using container support.

Re: AWS Is Asleep at the Lambda Wheel

#33
post #7

Python 3.11 is ~25% faster… Lambda is billed based on compute duration… if you are AWS PM Lead, it’s a challenging eng investment to justify.

It's actually pretty easy:

> Customer Obsession: Leaders start with the customer and work backwards. They work vigorously to earn and keep customer trust. Although leaders pay attention to competitors, they obsess over customers.

>Frugality: Accomplish more with less. Constraints breed resourcefulness, self-sufficiency, and invention. There are no extra points for growing headcount, budget size, or fixed expense.

> Deliver Results: Leaders focus on the key inputs for their business and deliver them with the right quality and in a timely fashion. Despite setbacks, they rise to the occasion and never settle.

https://amazon.jobs/content/en/our-workplace/leadership-prin...

Re: AWS Is Asleep at the Lambda Wheel

#34
Possible explanation: lambda just isn’t profitable enough.

Many workloads that are sporadic / rarely invoked gravitate to lambda.

Id wager that the cost it takes to build and deploy a lambda and manage its logs and security and storage are >> the revenue from invoking it 24 times per day.

Re: AWS Is Asleep at the Lambda Wheel

#35
Ran into the same issues with Node. 18 only took 7 months for them to support but 16 took almost a full year. For a paid service this is simply ridiculous. Maybe they can't be ready day 1 (though I sort of think they should have people following node/etc progress and be testing leading up to a release) but 11 months (heck, even 7) is way too long to wait.

This was made much more painful by the fact that M1 macs didn't really play super nice with node14 for whatever reason so my choices were 12 or nothing for a very long time.

Re: AWS Is Asleep at the Lambda Wheel

#36
post #4

This is a great article, and sums up my team's frustrations with AWS Lambda and the AWS engineering leadership in general. I worked in the past under a slew of former AWS product managers, and I know from them how the AWS roadmap works: 1.) Real, crucial customer feedback is incorporated into roadmap, usually for the core services that are already used extensively like Lambda. 2.) Executives have new, novel, sexy fla…

I mean has AWS ever been easy to use outside of S3? If they were simple, companies like Vercel and Heroku would not even exist (both runs on AWS infrastructure).

The complexity and amount of AWS-specific knowledge required to efficiently use it is pretty much absurd.

Even though AWS offers multiple different solutions (AWS SAM, Elastic Beanstalk, AWS CDK, AWS App Runner) that in theory should make consuming AWS services easier, these options still assume you understand and know how to use AWS.

In my previous job, I've been tasked to create an abstraction that will make using these AWS services easier and more developer-friendly.

I failed.

Yet I realized how hard it is, and how many people/companies are struggling with it.

That's why I've founded https://stacktape.com/. It makes using AWS simple to a point where any developer can efficiently use it. Let me know what you think.

Post reply on HN