Snapchat runs entirely on Google Cloud Platform, and have famously declared that they have 0 Ops people because everything they use is fully-managed and no-ops. This is one of the reasons they are able to deliver features at such a rapid pace - they aren't worried about scale or reliability the way anyone at the scale is on-premise or on AWS/Azure.
You're trading ops costs to be locked into a platform. Does that matter? Probably not if you're targeting being acquired by Google. If you intend to eventually move to your own equipment, it does matter.
There is no such thing as no ops. Only someone else doing ops, no more than I can have TensorFlow write code instead of developers.
Your fear is not invalid. That said, one can make the argument that speed of innovation trumps everything, and partnering with the best innovator is the most important thing when innovating yourself. This is akin to not wanting to ever date for fear of eventual breakup...
You make a good point on No-Ops vs Ops. I think it's a gradient. I also think that Snapchat is on the far far left (on No-Ops edge) against virtually any other large-scale technology company in the world, Google included (singe Google is Snapchat's Ops largely), largely because of the nature of Google AppEngine.
If you can map your workload and tasks onto Google Cloud Platform and the economics are in your favor, and you trust Google, it's a good thing. I'm sure it would work for many startups, but I'm curious how much of a lock-in it is and how much work it would be to move somewhere else.
Okay. What about their deployment CI/pipeline? I don't think GC offer that? I want to emphasize my philosophy on ops vs dev. I strongly encourage dev learn to do ops and ops should do dev work. So the question is, who is doing everything else outside of product development? Is dev doing all of the traditional ops-only work? Creating a self-service portal is both a dev and an ops job. It may just happen that someone is not 100% dedicated to do ops work. This is not even a debate of dev vs devops role. It would be disingenuous to say there is 0 Ops people if some dev are involved in doing Ops work (setting up DNS even programmatically). So I'd ask for clarification. Let's face it, someone is doing Ops work, but with a different title.
> the way anyone at the scale is on-premise or on AWS/Azure
Can you expand on this? How/why would anyone developing a potentially high-traffic app have to worry if they are e.g. leveraging AWS fully? How does that compare to google cloud?
Snapchat runs almost entirely on Google AppEngine. They deploy code and leverage APIs. Their millions of QPS and vast CPU, RAM, etc usage is managed fully by Google Cloud.
You can certainly replicate their scale on AWS/Azure.. or Google Compute Engine for that matter.. but you'd need to do a whole lot of non-trivial work around scalability, durability, etc etc. Snapchat runs light on this front, allowing them to focus on launching features rather than keeping the lights on. No one even remotely close to Snapchat in scale runs as lean as they do, and that luxury is afforded largely by Google AppEngine.
Snapchat runs entirely on Google Cloud Platform, and have famously declared that they have 0 Ops people because everything they use is fully-managed and no-ops. This is one of the reasons they are able to deliver features at such a rapid pace - they aren't worried about scale or reliability the way anyone at the scale is on-premise or on AWS/Azure.
reply