Great point and agreed, scaling down to zero and up from zero should be the default. I wonder, though, what exactly does this mean in the context of state. No matter if it’s an RDBMS or a NoSQL datastore, how does scale to zero look like? I mean, the point of a database is to persist your data. IOW: for stateless cases, I totally get it, for stateful ones, I’d appreciate if you could elaborate.
Excellent write-up, Kari! Thank you for taking the time. While I’m usually not a fan of shootouts or head-to-head comparisons, this piece really nailed it, very well done. If you ever plan to do a follow-up, I’d love to see Rust, PHP, Swift, C# in the mix as well ;)
Oh my, this is an anti-pattern and equivalent to (on a single *nix machine): “I give myself the UID 0, aka root, because I can’t be bothered to figure out what permissions exactly are required”. I mean’ just look at it in the docs. With this command, the current user, that is, you is made superuser or in other words:
Great post, thanks for putting this together! I’ve queued it for next week’s edition of my newsletter. Thanks also for including a link to our blog post, seems it needs some updating again.
Also, I’ll give a talk on this topic tomorrow at Velocity here in NYC using mainly examples from: http://troubleshooting.kubernetes.sh
It’s a marketing term trying to suggest there’s no operations involved. On the one hand there’s someone who has to provide the serverless offering (which might be outsourced to, say, Amazon in case of Lambda) and on the other hand someone needs to take care of things like monitoring, logging, troubleshooting, etc.—I shared some of my thoughts here: https://hackernoon.com/serverless-whos-on-call-now-81193fbfe9d7
Very nice job indeed! Also, interesting to see the overlap (esp. around namespaces) and differences when comparing to my post from earlier this week. We should really start to collect these kind of articles somewhere in a place where they’re easier to discovery ;)