In Fireship's last video, the guy talks about how Amazon moved Prime video's architecture to old-fashioned monolith, from Serverless, and saved a bunch of money and improved performance doing so.

It makes me wonder, what old tech have we put behind us already that will later make a comeback as an 'improved' version of that tech? What do you think?

Please, put me down as completely unsurprised.

"Nooooo! we need vertical and horizontal scalable MicroServices Arch! with lambdas! And kubernetes operators! Running on multiple cloud providers, across multiple regions!"
"Now listen here, you dipshit. The core can be implemented as 15 line of python script, using a cron to trigger every 6 hours. All that? not needed. gtfo."

vanilla web dev 😄

One should maybe mention the technical details behind it...

And read the case study.

Cause the obvious click bait "Microservice bad... Monolith good." Is just that. Click bait cringe.

https://primevideotech.com/video-st...

When one reads the whole article, I'd deduct another, very plain and logic truth: Don't do sth for the sole reason of doing it.

Don't apply microservices just cause a cargo cult tells you so or because it's hip.

Microservice architecture was a wrong architecture decision in this case. I'm bit surprised how they could have thought it was a good idea in the beginning, cause - even without an in depth understanding - one can spot the mistake pretty fast.

But there lies another pretty important lesson - and in my opinion the most important one... Be transparent, admit when sth is wrong and explain why.

Which they did.

Imho instead of going click bait cringy "Microservice bad, Monolith good" shit... One should more highlight that Amazon admitted and explained in detail a large scale architecture misdesign...

So others could LEARN from it.

And the lesson is - just for those tinfoil / vacuum heads who still don't get it - not "Microservice bad, monolith good"... But rather: "Observe, adapt, improve".

Cause that's what they did.

@IntrusionCM That's why it's important to learn not only how to do something, but also why it's done that way and how it works under the hood. Unfortunately with so many online courses and bootcamps, people only learn the former and skip all the "boring" parts.

So they realized that IO and process spawning aren't free?

@joewilliams007 Would be great. Let the server render forms again and only use js for some live validation like it used to be.

Is amazon applying the lessons learned from primevideo onto AWS? Would save a large amiunt of their resources!

Server side rendering and sttauc typing where those technologies for sure. OPs question what else could return once is even more interessting. Maybe the Goto as brand new blazingly fast programming paradigm called functionless? Or games on cartridges as new super save drm method? Newsgroups as maximum interopable decetralized social plattforn? Crt monitors as... I can't up with something for those. So most likely those.

@retoor Ooh, I wouldn't be validating with anything other than HTML5. That's one of the few developments that's certainly better - a simplification that works really well.

@max19931 That's actually one thing I find slightly surprising - why would Amazon want to expose the poor bang-per-buck/ROI of over-distributed systems? It's been the basis of AWS's business. It's almost what AWS does.

@spongegeoff i mean calculate the numbers but keep it internal and never publish the results. Just like exxonMobil did for Co2.