[sink]

As a former head of engineering for a mid-sized IoT company with many tens of thousands of devices deployed in the field using a competing product (no, not the AWS one), seeing this made my stomach churn.

This was my second biggest fear after waking up to a ransomware attack.

It's hard to imagine anyone trusting Google for IoT again. I will certainly put them at the bottom of my list for any other infrastructure I develop against in the future, and ensure that we have a documented exit strategy should it come to pass.

The idea of having just one year to develop against a new IoT core, test it, update all deployed devices, and then coordinate logistics and budget to do truck rolls when things invariably go wrong is really grinding my gears.

I feel for all of the startups having to deal with this. To the folks who are invariably scrambling, I really hope you either got advanced notice, or you're getting an extension far beyond what is publicized. Edit: The more I think about this, the more I want to believe there must be contracts in place for certain customers that extend the lifetime of this product beyond what is public. There must be.

IoT is not an easy business. Designing and programming hardware is hard. Supply chains are hard. Maintaining working inventory is hard. Building logistics networks for installation and maintenance is hard. Courting and explaining to investors why you don't have the profit margins of a pure SaaS business is hard. Relying on your cloud provider to give you more than 1 year notice should be the easy part.

[derefr]

> It's hard to imagine anyone trusting Google for IoT again.

I mean, presumably they decided the IoT vertical just wasn't for them altogether.

If Google intended some new service to take this one's place, they would have 1. launched it before deprecating this one; and 2. built a backend shim to route data sent to the old API into the new backend, so that people's code wouldn't have to change. (Like they did for Firebase, and for Stackdriver, and for anything else they actually cared about keeping the business of the customers of.)

This move, meanwhile, clearly sends the message of "we don't want to be in this business; stop trying to buy this kind of service from us; just go away." It's the feel of being on the receiving end of the "fire your [bad] customers" advice — just applied to "firing" an entire (bad?) market.

[dekhn]

Had I started an IoT company, I would not have depended on any cloud IoT product. VMs, blob storage, containers, load balancers, hosted databases. But not anything labelled "IoT".

[bigiain]

I have worked in an IoT company, and sometime time-to-market makes uncomfortable dependancies necessary. Hopefully people making those compromises have escape plans, but I doubt all of them do. I expect to read several "Our incredible journey" product end-of-life posts as Google kill off startups without a way to recover from this shutdown...

[speedgoose]

I worked at an IoT company, and I saw a few demos of IoT cloud solutions. We never found them to provide a lot of value. They were solving the easy problems and ignoring the hard ones.

[solarkraft]

> Relying on your cloud provider to give you more than 1 year notice should be the easy part

Isn't that exactly what contracts are for? Can you build an entire company on the premise that some random cloud service will always be available?

[senssei]

I fully agree that having more than just one year to migrate it's a bit short. Especially within the lifecycle of the devices that counts in decades.

Changing the ingress endpoint shouldn't be a big deal via OTA or configuration change. The lack of such configuration in the first place would be concerning for any new devices.

Ransomware attack it's a real threat.