Are we “stuck” with the ISO “stack”?
Saturday, November 24th, 2012 by Roberto Saracco
It was a long time ago, but I still remember it. We were in Melbourne, Australia, in the 80ies, and on a beach we were discussing how to decouple in an effective way the physical resources in a network from the “performances” needed by an application and we used a stick to sketch diagrams on the sand. It was a week-end during an ISO meeting that eventually resulted in the publication of the now “taken for granted” ISO stack.
But now we are “stuck” with it!
The decision taken at that time were based on the fact that an application in general would need many more resources than the ones a generic network could provide. And it would have been too costly to develop specific applications to fit specific network performances. Hence, the idea of decoupling one from the other. A set of layers would provide each a set of performances to the above layer, up to the application, making use of what the lower layer could provide and virtualising resources to provide more features to the layer above. So, as an example, you might have an application in Turin that needed to talk to another one that was located in New York and the physical network could not provide a direct link between the two. Well, an intermediate layer (transport) would provide a virtual link between Turin and New York by using the lower layer (network) to create a sequence of links from Turin to New York.
The system worked well, so well indeed that we are still using it, although today the TCP/IP has made some of these layers to collapse on one another.
The crucial implication, though, was, and still is, that an infrastructure Operator (a Telecom Operator) need only to be concerned with its own physical infrastructure and through inter-Operator agreement provide a trans-infrastructure path that could sustain end to end communications. But this end to end path was, and is, about physical links and transport. Whatever you are carrying on the pipes is not an Operator biz, it is an application biz.
As long as what the pipes were carrying was voice, and voice only, there was really no issue in terms of “transported value”. You could charge by the pipe: the longer the pipe, the longer time you were using them the more the price charged. But then the digitalisation of any media into bits has created a variety of different values at the application level that do not correspond to different values at the insfrastructure layer.
For a certain time this was not a big issue since resources at the infrastructure level remained relatively scarce and if an application needed more resources (e.g. to carry bit intensive information) it would have to request “more infrastructure” and hence to pay more. Than, almost all of a sudden, the infrastructure capacity multiplied by orders of magnitude (thanks to massive fibre deployment) and there were plenty of capacity for any kind of transport. This created a gap between the value perceived at the application layer and the value/cost of the bit transport that according to the ISO stack is independent of what is being transported. Almost at the same time, the digitalisation of “anything” made possible to transport voice and images, and video, and music and… you name it.
At the infrastructure level (layers) given the huge capacity all bits are grey. At the application layer bits have different colour, they generate different perception of value and therefore can be charged differently. This situation has opened up a Pandora vase of opportunities to enterprises that can exploit the cost/value differential existing between the infrastructure and the application layers.
Basically, you can get a 10 Mbps transport capacity from the network and use it to offer short message service. Charging by message, as Operator do (SMS) would charge 1,000 euros per MB. An application provider can squeeze into a 10Mbps stream 10,000 messages at a cost per message that is basically zero and a cost per MB of just, well almost zero too (if you like figures, an ADSL line with a 1 Mbps upstream capacity can carry 1.851 billion SMS per month, and an ADSL line with that upstream capacity is priced at about 50$ – hence the cost per SMS equivalent package is 2,7 millionth of a Eurocent! Clearly the real cost should take into account not just the transport but you get the gist!). This gap is therefore exploited by the Over The Top, OTT, with the result that the Operators see their revenue shrunk.
The customer benefits from this, clearly, the Operators lose revenues but this loss does not translate to significant revenues for the OTT since their cost is so low that they can afford it and get revenues from other sides, like ads. Hence the customer gains are not the result of a revenue shift but of a market depreciation!
This is why there is no point for an Operator to play the OTT game. They would not recoup the money!
Wrappin’ up: the ISO stack made sense at a time were resources were scarce, computation cost was high and could only be faced by centralising it, and all services were basically the same, voice, (more transport capacity required more resources and more could be charged).
Now, transport capacity is huge, and in most cases it exceeds demand, intelligence at the edges in the devices is more than enough to manage the infrastructure transport directly (layers from 4 to 7 are collapsed in the device/application) and we are rapidly moving towards the direct control of the physical layer (collapse of all the stack from layer 2 to layer 7 in the device/application), as an example a smartphone manages all the protocols and decides when to use a WiFi link and when to use a cellular radio link.
It is time for the Operators to change their approach to transport, acknowledging the flattening of the ISO stack and providing service directly to the application, end to end, even if this means leaving behind the concept of focussing on their own physical assets. They remain important but so are the immaterial assets of managing authentication, ownership, metadata and dynamically serving users needs based on her context and her specific needs. The times for one size fits all are over, if you want to make money!






