What's the difference between servers, serverless, and fluid compute?

Disingenuous to say serverless charges for idle time, when really you’re talking about AWS Lambda. Cloudflare only charges for CPU time, which really makes Serverless a lot more compelling. I know you’re trying to push “Fluid” compute but you should really take this down and redo

I think it's fair to associate serverless with it's most popular implementation in AWS Lambda. I explicitly said lambda in the video. You are correct there are other compute platforms out there, including Google Cloud Run, which I didn't cover in this short video.

I'm not trying to "push" but explain how fluid works. If you have more questions about that, I am happy to help.

That visualization is so killer!

What would be super sweet is if after you realise they using so much compute it's cheaper on a Dedicated server or vps you move them onto it and cut price that's real fluid.

Servers are the OGs. Serverless is like ordering instead of cooking. Fluid compute adapts like water

I like that

how'd you make the visualization? looks super slick

"Servers = Manual setup & fixed scaling. Serverless = Auto-scaling, pay-per-use. Fluid Compute = AI-driven, dynamic resource allocation across cloud & edge. #CloudComputing #Tech"

Wow this is an amazing way to visualize it

Great illustrations!

Could you open source the v0 prompt?

Dang you really killed it with explaining that honestly. I love hosting my projects on Vercal. One thing y'all definitely need to look into is better integrating the Supabase. The ability to move projects in Supabase from free to pro, when the pro is configured through Vercel. It wont let you move the projects around.

Nice! Great explainer. Time to try it out :)

This visualization is absolutely INSANE. I can't imagine how helpful this kind of visualization would be in video courses. I'm not a regular Vercel user, but I can appreciate where they stand with this level of attention to detail.

Super informative and easy to understand for most people. Great job guys !

cool cool, anyways

Did you watch the video? I talked about autoscaling with k8s.

The best ELI5 of Fluid compute

Lee strikes AGAIN 🙌🏻

awesome visualization

This model of "oh no, i over-provisioned i pay for usage, and oh no, i under-provisioned, my server went down." isn't very fair. The KBS topic you brought up is important, and autoscalers are still a real thing that companies you to scale for availability and cost efficiency

great vid

The visualization is so cool. Open source pls

@michael_rispoli 👀

1:10 lol

Is it fair to assume that fluid compute is much closer to a server model with extremely fast, managed horizontal scaling? Or can we call the underlying infrastructure for this something entirely new

It does feel closer to a server to me conceptually, especially with having 1 minimum active instance

Under high load, a server may slow down due to processing numerous requests and serverless scales by creating new instances for each request. Fluid will proactively detect instance overload and spin up new instance as needed for efficient request handling. Am I getting it right?

I tried redeploying a puppeteer based scraper application with fluid, but unfortunately the requests kept failing. It works fine with fluid compute disabled.

Won’t lambda continue to serve a few requests in that instance for the next X minutes before going idle?

Only one request is processed at a time, but yes if the function is already warm, then it doesn't need to spin up from zero to process the next request.

the difference these days is a few clicks or a few lines of TS (or JS if you're a barbarian)

should also show the number of requests processed next to usage

All this time I never knew what serverless is, and it turns out this model uses more server instances than server and fluid. The irony

Not to be offensive, I still don’t really get it. Ingress and load balancing are solved problems for Kubernetes. If you are not paying for cold start, that just likely means you are paying more per query. If you need to cold start, you probably shouldn’t be paying for your app because literally no one is using it.

You don't have to exaggerate, it's all just computers. 😅

Nothing new even cloudflare workers only charge for actual cpu time

The marketing team.

InterServer Has Been Providing Dedicated Hosting Services For The Past 25 Years.