Why the future of telecom will not be its past
When we first built telecom networks it was to enable voice over long distances. And we became a “thing”. An industry. An organisation. Organisations always come into existence to solve a need. Successful organisations then tend to grow very quickly. Then in many cases organisations forget why they exist and rather focus on self maintenance. And then, quite often, something or somebody arrives to disrupt the status quo and the incumbent rapidly declines, first in relevance and then finally, in existence. History has shown this to be true for empires, religions and more recently, Fortune 500 companies.
Let us apply this thinking to our industry — telecom. Let us pretend telecom doesn’t exist. Would we solve voice the same way? Do we need to? Would we start building networks the same way as currently do? Does the world need telecom? There are many alternatives on voice service now. There are wireless data alternatives. And many non telco companies are building networks in completely different ways with completely different architectures, economics and time to market. So why do we need telecom? What would telecom be if we were inventing it for the first time today?
Now is the time to ask such existential questions with absolute honesty since the world is going through large scale disruption and something else is required.
Seven milliseconds is how long the vestibulo-ocular reflex takes, the reflex that coordinates eye and head movements in the human body.
“The latency of action of the rotational vestibulo-ocular reflex (r-VOR) is 7–15 milliseconds, which is the time required for the eyes to respond in an equal, but opposite, manner to the motion of the head. This time is remarkably fast compared with the latency for visually mediated eye movements, which is longer than 75 milliseconds.” [ref]
If latency is larger then virtual reality becomes virtual drunkenness. It does not work. If we are outsourcing any of our reality to supporting digital infrastructure, that digital infrastructure has to ensure the signal can be sent, processed and the result returned, in less than 7 milliseconds.
Light travels approximately 300,000 km in one second so 300 km every millisecond.
There is approximately 510 million square kilometers of land on the earth’s surface. 33% is desert, 24% is mountainous, which leaves 43% habitable or approximately 219.3 million square kilometers.
If 1 millisecond latency is wanted between base station and “datacenter” then the base station is connected with optical and is positioned somewhere the area represented by a circle of radius 300 km, which is an area (pi * 300) = 942.5 square kilometers.
Therefore the total number of datacenters needed to provide 1 millisecond latency is one direction then one millisecond back is
219.3 million square kilometers / 942.5 square kilometers = 232,684 data centers globally.
Approximately 233,000 datacenters in the world can place compute and storage within one millisecond each way round trip from a base station, leaving 5 milliseconds for processing and enabling full neural interaction.
In 2013 there were already nearly 6 million mobile base stations deployed (macro only). So the conversion of less than 4% of the currently deployed base station sites into parallel compute and storage data centers provides enough coverage to enable global virtual reality for all.
Why is this a good solution?
Datacenter sites are already fully prepared, robust, powered for digital equipment. It is already the largest consistent distributed system on the planet. They are already connected, in many cases by fiber.
They are ready.
We call the final end state “future digital infrastructure” which is disaggregated and optical down to chipset — at birth.
What do we need at the edge?
Above is a global solution for solving the latency problem. This is not the only interesting aspect of the solution to consider however. In the following posts we shall speak about what capabilities are most valuable to have at the edge, when we are living in a data-centric real-time reality. We shall discuss what is required to build a fully distributed control system that enables a truly distributed high performance real time system.
And of course this is all built on the foundation of Software Defined Infrastructure — which is much more interesting than cloud.
For more on Software Defined Infrastructure and hyperscale see here.
For more on distributed cloud please see here
For more on the business potential of this architecture see here.
“When you make a choice, you change the future.”