I've gotten a few questions about Aurora Serverless v2 preview, so here's what I've learnt so far. Please feel free to chime in if I've missed anything important or got any of the facts wrong.

Alright, here goes the 🧵...

Q: does it replace the existing Aurora Serverless offering?
A: no, it lives side-by-side with the existing Aurora Serverless, which will still be available to you as "v1".
Q: Aurora Serverless v1 takes a few seconds to scale up, that's too much for our use case where we get a lot of spikes. Is that the same with v2?
A: no, v2 scales up in milliseconds, during preview the max ACU is only 32 though
Q: is the cold start for Aurora Serverless v2 still a few seconds?
A: yes, unfortunately...
Q: so if you want to avoid cold starts, what's the minimum ACU you have to run?
A: minimum ACU with v2 is 0.5
Q: does v2 still scale up in double increments, e.g. 4 ACU -> 8 ACU?
A: no, it scales up in increments of 0.5 ACUs, so it's a much tighter fit for your workload, so you'll waste less money on over-provisioned ACUs
Q: is there anything I can do with v2 that I can't do with v1?
A: yes, v2 supports all the Aurora features, including those that v1 is missing, such as global database, IAM auth and Lambda triggers
Q: wait, but it's twice as much per ACU!
A: yes, but v1 requires a lot of over-provisioning because it doubles ACU each time and takes 15 mins to scale down. v2 scales in 0.5 ACU increments and scales down in < 1 min. AND you get all the Aurora features!
Q: can you use "provisioned" and "serverless" instances in the same Aurora cluster?
A: yes you can! cool, right!?
Q: is data API supported on v2?
A: not in the preview, I'm guessing it'll be there in GA
Q: if using from Lambda, do I need to use RDS Proxy to manage the connections to the cluster? Data API kinda mitigated that for v1..
A: yes, you probably should, until data API is enabled on v2, otherwise, more connections = more ACUs, it can run you into trouble
did I miss anything? plz feel free to chimp in

@jeremy_daly has written a nice summary post on this too, with a nice experiment on the scaling behaviour of v2, so check it out if you haven't already https://t.co/DIGylN0HFX

More from Software

Kubernetes vs Serverless offerings

Why would you need Kubernetes when there are offerings like Vercel, Netlify, or AWS Lambda/Amplify that basically manage everything for you and offer even more?

Well, let's try to look at both approaches and draw our own conclusions!


1️⃣ A quick look at Kubernetes

Kubernetes is a container orchestrator and thus needs containers to begin with. It's a paradigm shift to more traditional software development, where components are developed, and then deployed to bare metal machines or VMs.

There are additional steps now: Making sure your application is suited to be containerized (12-factor apps, I look at you:
https://t.co/nuH4dmpUmf), containerizing the application, following some pretty well-proven standards, and then pushing the image to a registry.

After all that, you need to write specs which instruct Kubernetes what the desired state of your application is, and finally let Kubernetes do its work. It's certainly not a NoOps platform, as you'll still need people knowing what they do and how to handle Kubernetes.

2️⃣ A quick look at (some!) serverless offerings

The offer is pretty simple: You write the code, the platform handles everything else for you. It's basically leaning far to the NoOps side. There is not much to manage anymore.

Take your Next.js / Nuxt.js app, point the ...
Developer productivity, y'all. It is a three TRILLION dollar opportunity, per the stripe report.

Eng managers and directors, we have got to stop asking for "more headcount" and start treating this like the systems problem that it is. https://t.co/XJ0CkFdgiO

If you are getting barely more than 50% productivity out of your very expensive engineers, I can pretty much guarantee you cannot hire your way out of this resourcing issue. 😐

(the stripe report is here:

Say you've got a strategic initiative that 3 engineers to build and support it. Well, they're going to be swimming in the same muddy pipeline as everyone else at ~50%, so you're actually gotta source, hire and train 6, er make that 7 (gonna need another manager too now)...

...which actually understates the problem, because each person you add also adds friction and overhead to the system. Communication, coordination all get harder and processes get more complex and elaborate, etc.

So we could hire 7 people, or we could patch up our sociotechnical system to lose say only 25% productivity to tech debt, instead of 42%? 🤔

By my calculations, that would reclaim 3 engineers worth of capacity given a team of just 17-18 people.

You May Also Like

Still wondering about this 🤔

save as q
"I really want to break into Product Management"

make products.

"If only someone would tell me how I can get a startup to notice me."

Make Products.

"I guess it's impossible and I'll never break into the industry."


Courtesy of @edbrisson's wonderful thread on breaking into comics –
https://t.co/TgNblNSCBj – here is why the same applies to Product Management, too.

There is no better way of learning the craft of product, or proving your potential to employers, than just doing it.

You do not need anybody's permission. We don't have diplomas, nor doctorates. We can barely agree on a single standard of what a Product Manager is supposed to do.

But – there is at least one blindingly obvious industry consensus – a Product Manager makes Products.

And they don't need to be kept at the exact right temperature, given endless resource, or carefully protected in order to do this.

They find their own way.