Up unto this point, I have not interjected what I think very much. Because of that, I suspect that nobody has a clue of what I think. Here we go….I think that implementing services that have specialized traffic management for consumer services will drive ISPs out of business if they are done.
So, wait a minute I think you are saying. I thought the whole “Fast Lane” thing was going to be bad for the content folks. I think a broad implementation for this will be bad for the carriers. I have a lot of reasons for this and I will outline them here.
When I first got out of college, I worked for a company named Racal-Milgo. Milgo’s primary product was Leased Line Modems. These were used in IBM mainframe computer networks at places like Banks, Airlines, and Financial Services companies. To maximize the throughput of these networks, our customers used to get what were called “Conditioned Lines” from telcos. One of the big selling features for our products at Milgo was that we could measure whether the carriers were actually providing the service that the customer had ordered as a “Conditioned Line”. If the telco was not, the company got a refund.
If you are going to “sell” improved network performance, it is going to come with a Service Level Agreement (SLA). A company like Netflix is going to measure your network performance and want to get its extra money refunded if the SLA is not met. You can see Netflix already measuring and reporting problems in the Verizon Cease and Desist Letter.
In the consumer world, there are exactly 2 networks of scale: The Phone Network and The Internet. These two networks use the simplest forms of traffic management. Nobody has ever built a network that can run 10s of Millions of Endpoints with any other behavior. I don’t believe that a network can be implemented with very complex traffic management at that scale. Maybe such a network can be built in the distant future, but not today and not for a long time. It is not just a matter of compute power, it is a problem of network design.
Now to try to make a network that might be able to run the kind of service that a Netflix might buy into will require a substantial change in the equipment deployed throughout all layers of the network. This will cost a LOT more money than the current network AND the network will have to be deployed before a single dollar can be gained from a single potential customer. Between that and the way around dispute resolution, I think this is a game that can’t be won by the carriers.
So if this is TL:DR for you here are my arguments summarized:
– Specialized Connections have been around for over 30 years
– Even 30 years ago, Companies measured to make sure they were getting what they paid for
– Telcos failed to deliver even on the 1000s of such circuits that were static back then
– Companies were reimbursed for failure to meet contracts even then
– Why does anyone think that Telcos and Cablecos can now do this at the scale of 10s of Millons?
I think that people will find that this topic gets explored and never implemented once an actual conversation around an actual service and an actual SLA happens.
Jim Sackman
Focal Point Business Coaching
http://www.jimsackman.focalpointcoaching.com/
Change Your Business – Change Your Life!
Business Coaching, Sales Training, Web Marketing, Behavioral Assessments, Financial Analysis
Jim Sackman's Business Thoughts 
Jim,
You and I rarely disagree. This is an exception.
3002 Leased Lines isn’t an apt comparison, for a couple of reasons. From a technical perspective, a packet switched network is much more deterministic, not subject to environmental impairments and component aging, and easier to manage than a frequency division multiplexing network. From a business perspective, “special services” were a nuisance to a rate-of-return regulated telco; done right, assured flows can be profitable to a lightly regulated broadband provider.
The technical problems with QoS objectives and SLAs have been long since worked through. All the research in the 80s and 90s laid a solid foundation for rate conditioning, scheduling, admission control, etc. ATM demonstrated a viable architecture that could offer strong service guarantees. Diffserv and MPLS addressed the scaling problem. We’ve had SLAs for enterprise services for a long time. As far as I know, machinery for GPON/GEPON bandwidth allocation, Q-tagging, Diffserv building blocks, and managed LSPs, are implemented in every OLT, carrier-class switch and router currently deployed. Mostly, they’re realized in ASICS, ASSPs and/or network processors. If anything new is needed, it is going to be more scalable management… which could be an application for SDN/NFV.
And finally, SLAs will have to be edge-to-edge, not end-to-end (or end-to-edge). Existing machinery provides for performance measurement. And presumably, broadband operators aren’t going to make SLAs that they can’t support.
I envision a small number of assured services: probably one for ABR video, one for real-time streaming video, one for voice and perhaps one for M2M/IOT applications. I expect that consumers will be able to choose between one of these, at a premium, or Best Effort.
Dan,
I think you misunderstand my comparison. The reason for comparison was the point out that people who buy premium services measure to see that they are getting them.
Now the real reason we disagree is different. You have the wrong customer for the services. Let’s use Netflix as a straw man here. Consumer’s can’t purchase a premium service unless Netflix does. That means that Netflix is the customer and resells the premium nature of the service to the consumer.
Now follow along here. Netflix won’t buy anything unless it adds to their bottom line. None of the ones you have mentioned added anything to Netflix. They are going to want CBR service. The other problem this creates for the carrier is that they won’t be able to plan who wants to buy the premium service. The problem it creates for Netflix is that now they have to have the same service available from the major ISPs.
So, its a non-starter.