Hands up if you’re old enough to remember ATM here? And I don’t mean the type of ATM that sits on the side of a building dispensing cash – no I mean Asynchronous Transfer Mode.
For those who aren’t familiar with ATM, a little background. ATM was THE telco-grade packet-switching technology of choice for most carriers globally around the turn of the century. Who knows, there might still be some ATM switches/routers out there in the wild today.
ATM was a powerful beast, with enormous configurability and custom-designed with immense scale in mind. It was created by telco-grade standards bodies with the intent of carrying voice, video, data, whatever, over big data pipes.
With such pedigree, you may be wondering then, how it was beaten out by a technology that was designed to cheaply connect small groups of computers clustered within 100 metres of each other (and a theoretical maximum bandwidth of 10Mbps).
Why does the technology that scaled up to become carrier Ethernet exist in modern telco networks, whereas ATM is largely obsoleted? Others may beg to differ, and there are probably a multitude of factors, but I feel it boils down to operational simplicity. Customers wanted operational simplicity and operators didn’t want to have a degree in ATM just to be able to drive it. By being designed to be all things to all people (carriers), did that make ATM compromised from the start?
Now I’ll state up front that I love the initiative and collaboration being shown by many of the telcos in committing to open-source programs like ONAP. It’s a really exciting time for the industry. It’s a sign that the telcos are wresting control back from the vendors in terms of driving where the collective innovation goes.
Just like with ATM, are the big open source programs just too big and too complicated? Do you need a 100% focus on ONAP to be able to make it work, or even to follow all the moving parts? Are these initiatives trying to be all things to all carriers instead of changing needs to more simplified use cases?
“Sometimes the ‘right’ way to do it just doesn’t exist yet, but often it does exist but is very expensive. So, the question is whether the ‘cheap, bad’ solution gets better faster than the ‘expensive, good’ solution gets cheap. In the broader tech industry (as described in the ‘disruption’ concept), generally the cheap product gets good. The way that the PC grew and killed specialized professional hardware vendors like Sun and SGi is a good example. However, in mobile it has tended to be the other way around – the expensive good product gets cheaper faster than the cheap bad product can get good.”
Ben Evans here.
Is there an Ethernet equivalent in the OSS world, something that’s “cheap, bad” but getting better (and getting customer buy-in) rapidly?