Serverless was a big mistake... says Amazon

You're kinda using the terms microservices and serveless as if it was the same thing.

No shit!😂

the cloud = someone else's computer 🤫

vercel good 😄 lambda shirt 😩

I actually said in a job interview that even though IT is migrating towards microservices, I wouldn't discard the tendency to revert in the future. Like one day IT influencers might praise something like "no-latency apps" where there's no network calls, debugging is centralized, memory is shared, etc, and so we'd go back to monoliths.

bullshit

That fact Amazon switched from serverless back to monolith doesn’t mean serverless sucks. It just means they made wrong decision while choosing the solution for their case.

Beanstalk was a much nicer solution for the scalability needs of real world projects in my opinion.

NO solutions only tradeoffs - A Legend

Why does lambda have a reputation with lies..

Do and undo, the marching chant of IT.

ahahhahahahahhahahahhahahahahaahah

Is that equivalent to the difference in terms of intelligence between living beings based on central nervous systems VS those without a CNS?

Microservices are not synonymous with serverless. Certainly, the amount of data storage and computing power required by companies like Dropbox and Netflix, and the economic scales they can achieve, are quite different from those of an average company.

Totally disagree. There are far more benefits to a Serverless architecture compared to an old school provisioned server.

The last line with the Thomas Sowell reference can be a great way to finally show the far too leftist tech sector that.. they need to stop forcing something down everybody's throats when it doesn't even work for them in their life.

The video derive saved that money because they have a unique application which is more data transfer heavy than compute heavy. It’s not apples to apples for most use cases

Thomas Sowell ftw!

IIRC when I read the Prime Video blog a few months ago they explicitly state they were splitting videos into frames and processing each frame independently in separate lambdas. This is remarkably stupid any way you look at it - it's obvious I/O will kill you - and whichever architect/lead/jerkoff thought of it must have great selling skills or be cruising on past reputation because it's a ridiculous idea. So that means that their big problem with serverless doesn't generalize. (It may still be true for many current serverless users, but you have to look at each use case.) Meanwhile, contrary to the suggestion offered in this video, they're *not* inhibited from horizontal scaling due to the capacity of an individual machine. Processing separate videos is an "embarassingly parallel" problem - each video is independent of any other video - they spin up as many servers as they need each one processing an entire video at a time - not just one frame.

what people don't seem to get right with monolith vs microservices is ... that you have to choose the split points carefully. if you're dumb enough to split the middle of one logical flow into multiple microservices, slowdown will be the result. it you split let's say ... user account management, data querying, media serving into separate services, it'll be fine.

lol

Yeh configuration mgmt sucks for serverless. You probably just want workers or a pubsub topic or… I mean almost anything else.

"There are no solutions, only tradeoffs"
I died a little at this

This entire video is wrong. Amazon Prime moved from a microservices architecture to a monolith architecture. This has nothing to do with Serverless. You can still run both monolith and microservice architectures in physical servers (or ec2 instances).

Also an architecture does not become “bad” just because a tech giant decides to move away from it. If you read the blog they had a specific use case and monolith made sense for them. Does not mean it makes sense for everyone.

Before I get hate, I’m a software engineer at AWS and we choose microservices + monolith + serverless architecture based on the requirement.

As someone who works in the IT space and has to deal with the company wanting to "go to the cloud" I completely feel this video's message in my soul.

There's so many business-driven abstractions in modern software development that people often forget the fundamentals. Lambda is great but it's not magic - network calls are expensive, if you're moving the same data around hundreds of times a second you should be working in-memory

Years ago I was hired to manage the IT infrastructure of a company, then to help them move to the cloud to scale better and save money, then to install a cache server on location for speed and low latency. Now they are talking about leaving the cloud and buying servers. I am not in charge of making the decisions, but whoever is I just wanna say thank you because he is helping me pay for my mortgage with the same job 😂.

so you're telling me they saved 90% on server costs but are still charging me the same $9 a month?

I've been saying this from the get go: Using Amazon is the Most Expensive thing you can do. At today's server+hosting prices any company that does not have it's own infrastructure is truly run by incompetents! I've been saying this exact same thing of amazon when it was not "cool" to say so ...

I swear 90% of lambdas should be converted to K8s pod, I'm so sick of the extra hoops required to develop and debug lambdas. Please stop pitching these things, they take a lot more knowledge than a basic server.

I wonder if going monolith will affect Amazon’s e-commerce sector.

The first thing I learned about serverless in an AWS cost is that it is only useful for low cost operations, such as CRUD. Using it as a streaming service that calls another service is waaaay beyond the limit of what it was designed for.

sucks? not really. In the end, serverless is just a wrapper around exisiting system to allow creation a db service with much less code and infras prep than serverful. It does not suck, it's just far easier to create a db but far less flexible and not as capable compared serverful that's all. It's just a tradeoff. It probably sucks more right now because cost of hiring and maintaining programmers is at an all time low atm that's all.

I understood:
computers do things and these guys make the computers do things

Lol not a mistake for Amazon, as a provider

I'm the kind who has been avoiding AWS and other cloud solution for 10+ years with a passion. Costs more at the start but you always end up saving pretty quickly... Also true start ups, and by that I mean not ones with VC funds, rarely actually use cloud solutions in my experience.

The Duh Heard Round The World

Serverless means AWS does the server patching instead of you.

Never trust an external company to architect your solutions. They can build it, or provide components, but you need to architect. The issue with technology focused architects, from AWS/Azure, etc. is that they believe that everything must run the way they are told to run, i.e. use an architect playbook.

So yeah - pick technologies, not architecture.

No shit uploading frames to an S3 bucket and splitting that kind of thing up is a bad idea LMAO

TL/DR: don't look at big tech exp, they don't use bleeding edge solutions and usually lose when it comes to overcheck for how much money they would loose. Serverless and microservices are fine, but they got their own usecases and big tech as well as a lot of not so intelligent devs use any hyped technology with no any decent reason, so they get what they deserve

I worked at VK sometimes ago, it's a huge big tech (over 30k stuff, a couple of thousand of teams). And even thou they use microservices and advanced devops, their approaches are kinda deprecated cause big tech can't try all new features released within last 5 years. So such big techs usually get the majority of big numbers of outcome that blows readers mind so they tryna follow their voice even thou big tech just sucks and often spends money in extremely non optimized way cause... they are just don't have time for it!
Just recall gamedev industry and how AAA companies spend colossal amounts of money on dev and ads, how their devs create the worst codebases in the whole IT industry cause they don't have time and gotta make everything's done asap, how they work for 12-16 hours a day, and how bad and bugful they projects are even tho they delay their releases. Or how they create the same game every year calling it as a new part? But newbies in industry unfortunately look at big tech and gamedev and learn from them...

I am now on 3rd semester on my computer science college course, I have little idea what this video means

why do amazon charge themselves tho?

Me pretending to understand what is going on

Scaling only matters if the demand is bigger than you planned for. The vast majority of applications do not suffer from this, either because they are big enough that they planned for it or they are tiny and will remain so. Small things that unexpectedly turn into big things are extremely rare.

This is why there is no 1 architecture fits all and it is case by case.

It's not surprising that the biggest cloud supplier isn't happy about customers not depending on their platform... why is it an interesting statement though?
The whole point of serverless is that you shouldn't depend on lock in for your applications.
Your applications should be able to run independent of platform utilized for the workload.
That is terrible for Amazon's dominance, for sure.