Information utility is all about ubiquitous access to data, which is all about the network, which is all about the balance of what is in the network versus what is on the network, and more explicitly, how that balance is managed – or operationalized, in the lingo of those who deal with such things.
Network Operations is a discipline in its infancy, but one which begs for structure, maturity, and automation. Once we can reliably extract high quality (repeatable, predictable, profitable) operations of what is in the network, we can then extract the same from what is on the network.
This grid operation will finally propel the network common carrier from monopolist to utility – with all the concurrent regulatory structure and commoditization – and the network service and content provider from pawn broker to banker – with all of that concurrent regulatory structure and commoditization – and it is the combined delivery of these two commodities from central sources in set volumes and priced on consumption which creates Information Utility.
The grid in the title refers to the interconnectivity between those served by the utility at the edge and those providing the utility at the core. This build-out, while continuous, is largely complete for all but the least dense population zones and with it we can deliver reliable, high capacity links to individual households. The proliferation of mobile data networks and the migration to the IPv6 will increase this granularity to the individual device.
The box and the meter, however, remain to be properly deployed.
The box is a reference to the shipping container and the way that it transformed not only the practice of shipping, but also the economics of shipping. Prior to the shipping container, cargo was billed by the content, and shipped break-bulk, that is hand packed and hand loaded. After the shipping container, cargo was billed by container, and by weight; what the cargo was no longer mattered.
Similarly, the electric meters and gas station pumps are consumption billing, and you are billed for what you use, not for an arbitrary allotment.
Initially, it is easy to think that these two – the box and the meter – are somehow exclusive, but really the box is what the meter is counting – the kilowatt or gallon – over time, as the flow passes by.
So, what are good boxes in network operations?
In the network there are nice parcels that can be counted easily – packets or bytes – and are already built into every piece of networking equipment. On the network, the question becomes a little more difficult. What ought you be billed for, who should be billing you for it, how do you keep from getting billed twice (or more) for it? These are the relevant questions.
For clarity, by “on the network” I mean that which happens above Layer 7 of the OSI model – http is in the network, html is not. So that tends to steer me towards countables such as number of concurrent conversations or the quality of service being used. Ultimately, what we need is not a byzantine formula itemized on every bill, but a single unit of measure that can be applied with equal voracity to a single fat connection or many thin ones. What we need is an Ohm’s Law for data transfers, and from that, a unit of power analogous to the Watt.
I say this and do not say that we need a unit analogous to the gallon intentionally, because I look at GiB or GiB/Sec as insufficient for dealing with the total cost of connectivity in much the same way that gallons or gallons/mile does not sufficiently account for the cost of the upkeep of the highway system or lend itself to dealing with congestion.
So we need a unit derived from the following:
- packets (per second)
- bytes (per second)
- concurrency
- quality
And we need that unit to have some sort of intrinsic validity much like the relation between voltage, current, and impedance have with power.
What about something like this:
(bytes/concurrency)*packets = quality
What kind of metric would that give us? I don’t know that it really gets us a Law, since the quality value appears to be arbitrary and not present in the current protocols in such a way that is mathematically constant. On the other hand, we do have the ability to arrive at quality easily enough to investigate a posteri whether or not such a relationship is sufficient to derive standards that can be applied back into the network standards.
I also don’t think of this as purely escaping the “byzantine formula” which I was trying to avoid, inasmuch as we are not familiar with the outcome and don’t have a name for the unit of measure it would represent. It does give us the possibility of finding a box to use as the metered value that is more fully representative of the actual forces at play when delivering data from core to edge, and if not the final answer, it does appear to orient us in the right direction for such a box.