The problem is that there are now multiple generations of software engineers that do not know how bandwidth is priced. They've only used managed providers that charge per unit of ingress/egress, at some fractional dollar per GB.
I show them bare metal providers and colo that bills by size of pipe rather than transfer. They refuse to believe it or assume there must be a catch. There usually isn’t, though sometimes the very cheapest skimp on things like rich peering and can be slightly slower or less reliable. But still cheapest is relative here. Expensive bare metal or colo bandwidth is still usually hundreds of times less than big three cloud egress.
It’s just nuts.
It’s a subset of a wider problem of multiple generations of developers being fully brainwashed by “cloud native” in lots of ways. What an amazing racket this all has been for providers…
$0.09/GB? I guess that’s just what it costs. Expected when you’ve never looked at, considered, or even heard of things like peering and buying transit or even co-location. Enter 95th percentile billing on a gig port for $500/mo or whatever…
Same goes for hardware. Want to watch something glorious? Take someone who has only ever used VMs, etc in big cloud and give them even a moderate spec new bare metal server. The performance melts their brains.
Then they realize the entire thing - colo, servers, bandwidth, even staff is a fraction of the cost.
The industry really found a way to stop the downward cost spiral of personal computing and bring everyone back into mainframe. Moore’s Law is not dead but you no longer see the benefits. The difference is just pocketed by cloud providers who will use the gains to pack more low performing VMs onto less hardware.
5c/GB is less than the 9c that Amazon charges, but still, ouch.
You buy space in Equinix DCs because they are network neutral and have a high density of ISP POPs, i.e you can light up fibre between your cage and HE, Level3, whatever other transits you want to use + peer with everyone on the local peering fabric super easily.
I consider Equinix transit to be an option only when you really can't be assed to do anything else and you don't intend to hammer the links anyway, otherwise yeah you should be doing all the things above because otherwise you could have rented colo in some shitty backwater DC instead.
You pay hurricane for the bandwidth.
No 1 believes they can do anything anymore. It's not just hosting and hardware either. Everything single thing becomes a SAAS that gets purchased and they all refuse to believe it can be done in-house.
> It’s just nuts.
It's exactly that. I've seen upper management in many companies value external solutions as more cost effective because if they had to do it internally it'd cost 2x as much.