Wait, is cloud bad?
newsletter.goodtechthings.com
newsletter.goodtechthings.com
A lot has changed since AWS/EC2 was launched. Almost 2 decades have passed. It is time to refine our thinking if we are still stuck in the world of VMs and containers. I think the next big wave is "cloud native", but I have detected some problems. Between all of the partners, vendors & customers I work with, I'm operating with 5+ totally unique & mutually-exclusive definitions of what this term actually means.
The harsh reality that I've arrived at is that an aggressive application of first-party-only-actually-cloud-native-for-real will result in the loss of 80%+ of existing IT/dev staff in most big org charts. I've been on several consulting calls this month alone that indicate much of the ongoing on-prem+cloud hybrid clown shows are simply an ass-covering exercise. We showed a truly cloud native SaaS app ecosystem to a prospect and their developers flipped into a very defensive posture within seconds of realizing how few resources we used to eclipse their in-house efforts.
To be clear - I still don't think cloud is for everyone, but if you are going to do it half-assed, I won't bother considering the statistical relevance of your attempt.
It’s not. Traditional hosting, data center relationships, and “server in the closet” still give competitive advantage to organizations that use them wisely.
There’s really not much else to say that isn’t just meaningless rambling about imagined orgs/projects that better run here or there.
What most orgs need is just an astute advisor who isn’t sold on any one solution and can identify the right fit. And of they can’t afford that, they’re probably fine on a cheap VPS. :)
However you have to really know what you are buying.
Some stuff; cloud computing is exceptional.
Want to prototype a full deployment? want to get it into production?
The defaults are pretty good (in GCP at least) and can lead you to a highly reliable system pretty easily.
However. On the whole cloud is anywhere from 5-11x the cost on paper. Operationally it might save you half the time assuming you have a good dedicated hosting provider nearby- and you don’t need to reinvent some primitives such as blob storage.
Theres a lot of discussion surrounding what I refer to as “cloud myths”. In every thread where cloud is mentioned someone will say one of the following:
* Headcount savings
* competence savings
* near-infinite scalability.
Unfortunately you need specific competences for cloud (most effort is on the OS level anyway even with colo so all you have done is add the burden of abstraction and it is a lot of abstraction), headcounts are comparable too- especially when you talk about finops, devops (terraform etc). Same too for infinite scale. I have looked down the barrel at how scalable cloud is- unless you're giving a couple months heads up, if your compute use is high you can easily exhaust a region or zone.
If you are happy to spend 5x more to get where you're going faster then cloud is pretty great. But my advice is to have a plan to migrate to something cheaper once you figure out your baseline.
I think you are wrong about the finops stuff
Everybody should do finops, this is not a cloud-specific skills .. At least on compagnies I worked for, there were no finops in non-cloud infrastructure, there were no cost reports, nobody knew how much things costed
Which is aweful Just because you can do finops on public cloud does not mean you must: after all, you can stay on your good old crap
Or skills, which is why I've noticed government departments are migrating everything to public cloud, no matter how expensive it is.
Fundamentally, they just cannot hire skilled staff, because they're legally prohibited from paying the going market rates. Those skilled staff also tend to stay away because they're put off by the extreme bureaucracy and generally unpleasant work environment.
Anything that can be offloaded to a non-staff contract is pounced upon.