Internet in a box or craving for a better performance

I came across the site, which promises to create a small device with a local copy of a terabyte of internet's free information. The purpose of this project is to make a wealth of internet resources available in environments without a connection to the internet, particularly, the schools.

I was reminded of this project as I waited patiently for a streaming video to buffer. This was even after getting a higher bandwidth net connection. You would not expect “Internet in box” to resolve the performance problem if you install this device. But what if an ISP used such a device, would your experience of streaming video be better?

Would the experience of Stockholm's efforts at traffic congestion be at all relevant? Stockholm introduced a charge for vehicles entering the central city and improved the commuting time. The results are explained in a very nice talk by Jonas Eliasson on

It wasn't nudging people to modify human behaviour, though an excellent lesson, that was relevant. Congestion is a non-linear phenomenon and would be equally relevant for latency delays. Hence, the relevant lesson is that a small decrease in traffic can have a substatial impact on congestion.

Which means that we are now in a position where we have reduced traffic across this toll cordon with 20 percent, and reduced congestion by enormous numbers, and people aren't even aware that they have changed, and they honestly believe that they have liked this all along.

Open Source Downloads

Your current experience may have been that when updating a linux distribution, the speeds are better if the servers are in the US! There are hardly any mirrors in India. Shortest/fastest data path to a mirror in the neighbouring countries might well be via the US.

An ISP can legally and morally host or mirror such opensource content. All the open source distributions encourage creation of new mirrors. An ISP can deliver all requests for such content from its local copy. Unfortunately, an ISP may feel that the volume of traffic is not that high for becoming a mirror to be worth the cost and effort.

But, the experience of Stockholm traffic could be a guideline. The realisation that the congestion is non-linear may imply a far greater impact on latency than the quantum of open source download trafic might indicate.

If local mirroring helps, then all the ISP customers benefit. Each one will experience fewer instances of degraded performance, especially when dealing with streaming content. Of course, as a reader of this magazine, you will benefit more because you are likely to be a consumer of open source content.

The ISP will gain as well. The first gain is that an ISP's external bandwidth requirement may be lower for the same level of performance. The second gain is that their revenues will increase if my personal experience is anything to go by. I, in fact, reduced my bandwidth subscription because of latency and couldn't consume the bandwidth I had wanted to use.

You may also have ignored viewing a lot of content because it is irritating to watch it in a start and stop mode. Furthermore, time is a limiting quantity. If you use more time to watch one video, you have less time for additional content. An obvious possibility is that if you could view more, you would be willing to subscibe for a larger bandwidth and pay more to the ISP.

It would have been great to simulate the overall impact of downloading from a local mirror in contrast to downloading from an external mirror. Unfortunately, this simulation needs the data and the resources of an ISP to do it. It would be nice if there were a way to nudge an ISP.

An alternate hope for better internet performance is Remy, a project by MIT, that generates TCP congestion-control algorithms which “significantly outperformed algorithms devised by human engineers”(A faster Internet — designed by computers? - MIT news)

There is hope. It wasn't long ago that you may have disabled images to get reasonable browsing experience. Today, you are able to view a fair number of videos. If past is anything to go by, you can be certain that the performance next year will be better than it is today. The wish, though, is that why can't the improvements be visible tomorrow!

Aug 2013