Hacker News Reader

Comment by BizarreByte | Hacker News Reader

I worked on a project that did this very thing to determine if the software was online. It’s a very common way of doing it, even if it’s not the best way, because it’s reliable, simple, and cheap.

View on HN

Well, just connect to your own service, then? If it's offline, then you will not be able to "sync data" or whatever anyway, even if the device is connected to "The Internet". These connectivity checks sound completely pointless.

The devices would reach out to other third party services as well. The check would ensure they didn’t pointlessly do so. Knowing whether there was internet or not was used for a lot more than talking to our own servers.

> Well, just connect to your own service, then?

When you’re pinching pennies as a startup there’s a real cost associated with that. Several thousand devices checking in every few minutes isn’t expensive, but abusing Google was free.

Fwiw I’m not defending this, I’m just saying it’s common even if bad practice. There’s a reason IoT devices have a bad reputation.

Why pay for bandwidth when you can make Google take the cost?

I don’t agree with it but I’m sure it saves a few dollars.

You don't pay anything. You either complete whatever you intended, or if there's no connection to your server, no request hits your server.

It doesn't save anything.

It is always better for debugging purposes to be able to distinguish if the problem is with the client's connection (if google/baidu/yandex all are not reachable) or with the vendor server (if those former are reachable but the vendor server isn't).

Then ping your own server, and if there’s a failure then fallback to pining another public service with known high availability to determine if it’s an Internet issue. Pinging Yandex by default every 5 minutes is not necessary

Do you know how many ovens there are? The AWS lambda fees are going to be insane!

So much cheaper just to ping google.com instead.

I worked for a Company that was a vendor for Lenovo. They had some sort of "News about Lenovo" app on all of their commercial grade machines. The content and translations were managed in our system and each computer would pull down the feed from our servers. It was decent traffic, but nothing crazy... Until they made a change and all of the sudden every machine was pulling every hour at zero minutes and zero seconds. We had one minute of DDoS every day at noon for about a week until that was mitigated.

Hopefully no one is using lambda for connection health checks. An empty static page on a CDN is probably a better use of resources.

But yeah, at that point just ping google since they’ve spent millions on availability.

Let's see: ping packet size is say 64 bytes so 128B both ways

My $3 VPS has 1TiB/mo bandwidth limit. So that's 2^40 / 128 / 24 / 60 / 30 = ~200k ovens pinging once every minute (which is ridiculous frequency) for $3

I guess if manufacturer can't afford $3 per 200k ovens sold, he should go out of business or stop stuffing ovens with IoT shit. :)

I can't imagine this passing code review on any project I've worked on.

It was a startup, things weren’t super strict.

Reply on news.ycombinator.com