Looks like https://deno.com/deploy will be a managed service - the implication seems to be that the default option will be to use their CDN to run code with an option to DIY if you prefer.
How does this business model survive Amazon AWS making a blog post, "Here's a template to run your deno code on Lambda!"? They'll never beat AWS on costs in the long term. They can burn VC cash to stay afloat and try I guess.
AWS sucks hairy balls at providing things that are simple for developers to use, so that could be their competitive advantage, but I'm just guessing here.
I find the idea that AWS will just eat the competition always a little silly as well. I've used AWS managed offerings that were far inferior to the alternatives.
Usually because the company already pays an AWS bill. No one will care if you add another Lambda function, but to use an alternative you’ll have to get past some gatekeepers.
And yet we've never seen anyone beat Amazon yet. Look at how much the PaaS area has churned over the last 10 years. Giants that stuck around like Docker are completely deflated and near worthless compared to their initial values and expectations. The smart ones like Heroku got out when the getting was good.
> And yet we've never seen anyone beat Amazon yet.
That largely depends on your definitions of "beat" and "win". There are plenty of software infrastructure firms out there that Amazon has yet to smash into the ground.
Nature seems to think (and I agree) simply existing is winning.
Lambda uses containers vs. cloudflare workers use v8 isolates. v8 Isolates are much much faster and more secure for serverless functions.
Deno seems to be targeting cloudflare as a competitor for their service... But it's probable that AWS will release a cloudflare worker competitor themselves if deno continues with the MIT license.
> Lambda uses containers vs. cloudflare workers use v8 isolates. v8 Isolates are much much faster and more secure for serverless functions.
You're right that v8 Isolates are blazing fast, but Lambda runs functions in a microvm spawn by Firecracker [0], which is likely to be more, not less, secure than Isolates [1].
> Firecracker [0], likely to be more, not less, secure than Isolates [1]
This is debatable. It's true that V8 is a much larger attack surface than Firecracker, therefore likely to have more security bugs than Firecracker itself. However, Firecracker runs attacker-provided native code directly on hardware, which means that hardware itself becomes an attack surface, one that is quite wide, not fully documented, and very hard to patch if problems arise. It's much easier to work around hardware bugs when you're working from JS / Wasm and can control the code generation.
Ultimately I don't think you can really say one or the other model is more or less secure.
(Disclosure: I'm the tech lead for Cloudflare Workers so I am obviously biased here.)
> Firecracker runs attacker-provided native code directly on hardware, which means that hardware itself becomes an attack surface, one that is quite wide, not fully documented, and very hard to patch if problems arise. It's much easier to work around hardware bugs...
I see your point. I mean, Google wouldn't put as much effort as they are on gVisor if KVMs were the best possible answer.
To be fair, gVisor also runs native code directly on hardware. Any modern VM-based system is still depending on the CPU to enforce boundaries. A big CPU bug could ruin that at any time. (Spectre has been pretty bad, but not quite a showstopper...)
That’s the goal of almost all startups. Your chances of being acquired by a FAANG company are millions of times greater than going IPO. It is almost always the exit strategy.
Workers is now a much capable platform. It supports eventual KV storage, caching large files, longer runtimes (30m+), WebSockets, WASM-executables, and distributed mail-boxes àla Erlang.
Lambda integrates with existing AWS services, whilst Cloudflare invents newer services to go along with the serverless-first paradigm. Different strategies but they do compete with each other.
Eh, Vercel pulled it off rather well with Next.js. Given that Guillermo Rauch is an investor in Deno, I wouldn't be surprised if they partnered in some way.
