5 Dec 2014

Interesting Ideas from Internet.org

Mark Zucerberg wrote a fascinating paper regarding Facebook’s efforts with Internet.org to bring the Internet to everyone. This paper is full of fascinating information [1] and insights into this problem.


Mark explains that he’s doing this as a social good, and not as a profit center. This is because people who have little money to spend are not an attractive audience for an ad-supported site — they won’t generate much revenue for years, if ever.


To begin with, Mark points out smartphone sales are exploding worldwide, but data plans are being adopted far more slowly. Which means that many smartphone users won’t have Internet access. This is a highly counter-intuitive situation for those of lucky enough to be able to afford mobile data.


To fix this, the Internet needs to be either free, or very cheap. In the Indian context, I’d guess that means something like ₹10 per month.


But running the Internet costs tens of billions of dollars annually. Licensing spectrum alone costs tens of billions of dollars a year [2], to say nothing of wired networking, land, electricity, and so on.


And all the companies involved need to make a profit to stay in business, so the only way we can reduce the price people pay for Internet is to reduce the cost of running it.


Mark is aiming for a 100x reduction in the cost of accessing the Internet. This consists of a 10x reduction in the amount of data a device consumes, and another 10x reduction in the cost to deliver a bit to the user.


Partly, this is about using data more efficiently. Mark mentions how Facebook optimised their Android app from using 8-10MB per day down to only around 1MB per day. Most apps are optimised for latency, responsiveness and battery life, not for the amount of data they consume. This is because the engineers and other decision-makers are from the developed world, and they have fast, low-latency, high-bandwidth, unlimited connections at their disposal.


Mark envisions a free or ultra-cheap tier of Internet that is mostly text, with no video, audio, or photos. There would be no apps, because users wouldn’t have the bandwidth to download megabytes for each app binary. It would just be text rendered by servers. This tier would be cheap enough to be subsidised by the government or other entities, or for poor people to pay for out of their own pocket, like ₹10 per month. This tier would provide access to a basic set of services like Facebook, search, Wikipedia, etc. Users who want more can pay. If we get this differentiation right, Mark explains, we can provide free or cheap Internet to most people in the world, while generating enough money to keep the Internet running sustainably.


Then, Mark mentions different technologies to serve different densities of users, from cell towers that serve maybe less than a km, to drones and balloons that maybe serve a city, to low-earth satellites that serve a bigger region, to geosynchronous satellites that can serve an entire continent. As you go up, you can serve more people, but since the combined bandwidth of the satellite is limited, it makes more sense for rural areas.


Mark mentions Facebook’s app for feature phones, and how it loses 20% of its users every month (because they buy smartphones), but more people who didn’t have Internet on their phone or didn’t use Facebook sign up, so the feature phone app’s usage is actually growing. This is an interesting situation. I can’t think of any other product or service that loses 20% of its users every month but is still growing.


Finally, to make the Internet cheap, we need to make servers and datacenters cheap, as well. Every device imposes a load on the cloud, on all the services it syncs with or otherwise talks to. This is perhaps obvious in retrospect, but it was not at all obvious to me before I read the paper. Facebook’s Open Compute is important here.


All in all, these are fascinating insights from Mark into bringing the Internet to everyone. It’s impressive to see his clear, analytical thinking, and how he boils down a complex problem to its root causes, which he then comes up with a plan for attacking. If you want to solve a problem, but you don’t know what its root causes are, you’ll fail. Mark’s clarity of thinking is impressive.


[1] There are also interesting statistics: 2.7 billion people are online. There are 5 billion phones in the world. 1 billion of them are smartphones.


[2] This is an argument for open spectrum.

No comments:

Post a Comment