Why Can't The Network Be Smart Too?
from the battle-for-the-smarts dept
For years, David Isenberg has been pushing the idea that the network going to be stupid, with all the smarts at the end. He actually wrote this while working at AT&T, but left soon afterwards. For years, it looked like AT&T and other telcos missed the point, and continued to assume that all the smarts would be at the center of the network, like the old telephone system. However, as AT&T has stepped into the VoIP world it appears they’re at least realizing that distributed smarts is the way the world works. However, they’re not all the way there. Some have been complaining that many of the VoIP providers haven’t been willing to open up and let others develop intelligent devices to make their networks even better, which is just the old centralized telephone network mindset at work. In an interview with Business Week, AT&T’s chief technology officer, chief information officer, and president of its Global Networking Technology Services business, Hossein Eslambolchi, tries to position the AT&T position as being a “middle ground” by allowing some smarts at the end, while keeping more smarts in the middle. This isn’t so much a compromise solution, because the smarts at the end are already there and growing. It’s more like an effort to try to cling to some of the smarts in the middle, because that’s the only place they can charge for things.
Comments on “Why Can't The Network Be Smart Too?”
Difficult business decisions
The thing about the telcos is that they already have a massive network that is based on having the “smarts in the middle,” so to speak. For them to simply scrap that and start from scratch on a “dumber” network wouldn’t necessarily be the wisest thing to do. It’s not (entirely) a matter of what they can and can’t charge for; one must also take into account the costs of changing their entire infrastructure. Changes seem to be in order, but we’re more likely to see a gradual phase-out than a willy-nilly apeshit all-out conversion, taking place in the space of 6 months and gutting everything they have…
Marketing gibberish
The carriers have to claim that the network must be smart in order to attempt to pry themselves out of a commodity business.
The fact is, you want the net to be as “dumb” as possible, where “dumb” is defined as “making as few assumptions about the nature of the traffic as possible.”
Clearly the implementation of the net is pretty hairy, with complex routers doing a surprising amount of computation. But it’s the same computation whether your protocol is VOIP, HTTP, SMTP, or bittorrent. Had the network been a standard “smart” net in the sense that the phone companies use it, then we would have had only SMTP, FTP, TELNET, etc….for all the other ones I mention above (and a gajillion others) had not been even imagined back in 1984 at the time of the NCP->TCP transition that laid the technical foundation for the ARPAnet->Internet transition.
I think the carriers are doomed and that packet transport will end up being like the streets — a largely unregulated and unmetered commons with some premium transport channels for those who think it worthwhile.
Dumb is good!
The dumb network is scalable.
If the smarts are in the network then the network has a bottleneck – a vulnerable point.
This dumb network has:-
1. Pages and files.
2. Indexes – aka Google etc.
3. Realtime searchers – blogdex, feedster etc.
4. RSS feeds.
And ??? – what ever will come tomorrow.
All have a role to play.
The rapid growth of new services occurs because there is no central control.
Google provided a service that was needed – but at the time didn’t seem valuable!
The next step will occur by seeing a way of combining the current parts with a new idea.
What is it to be?