Welcome to lecture fifteen of Networks, Friends, Money and Bytes. And today we would talk about a peer-to-peer networks, such as Skype for voice communication and BitTorrent for file sharing and video streaming. And this starts a sequence of several lectures on the challenges faced in technological networks. One big challenge is the massive amount of content distribution. The content could be file, Could be video, And we'll look at peer-to-peer and cloud or data center technologies as solutions for this problem. And another problem is a prevalent adoption of mobile wireless technologies of various kinds, And we would be talking about this in lecture eighteen. So, today, we'll focus on peer-to-peer for massive amount of content distribution and we will look at two themes. One is the theme on overlay networks. We will see the power of building overlay networks based on basic connectivity services offered by the internet, by IP networks, providing simple, stable, ubiquitous, and economical connectivity. The second theme today is how to scale up a system? We'll look at a different kind of scaling up method in the next lecture for cloud and content, for cloud and the data center. And today we'll look at how a peer-to-peer provide an intrinsically scalable solution. Now P2P is a technology that's been developing for more than fourteen years. Back in the late 90s, around 98 to 99, Kazaa and Gnutella were provided and they became very, very popular around 90, 1999 to 2001. But they also face a few problems. One of which is copyright issues, cuz a lot of content on these P2P networks violated copyrights. Another is the issue of free riders. A lot of users do not contribute the content or bandwidth and yet they consume from others. These are the, if you will, first generation of P2P networks on top of Internet. And then, in the same year, 2001, two major companies came to existence. One is Skype. Okay? It started from Kazaa, was equired by eBay in 2006, for 2.6 billion and then it was, again, acquired by Microsoft a few years later in 2011 for, this time, over $eight billion And today, there are over 700 million users worldwide using Skype, And it's reported that on average, every single day, there's over 700 million minutes of Skype calls. Now, another major company that took place in the same year is BitTorrent. Okay? It is also a P2P technology, but this time it is for content sharing and, it turns out that half of the Internet traffic in the midnight 2000, Around 2005 to 2007, Was estimated that nearly half of the all the Internet traffic were using BitTorrent protocol, so there was a huge success of this technology. Now, there are some major differences between Skype and BitTorrent. Both are free, okay? They are free services, unless you need to, need to, need to use, Skype to call a normal phone, landline phones or you want to use a certain kind of conference call facilities. Otherwise, Skype is free, and BitTorrent is free to install, and furthermore, the content on BitTorrent is also free. So how could they be free? Well, Skype uses P2P for signaling. Now, including locating each other, including establishing connections. Okay. So this is a so-called control plane use of P2P technology. In contrast, BitTorrent uses P2P for actual content sharing in the data plane. Later in lecture nineteen, we'll talk more about control versus data plane. Roughly speaking, you can think of data plane as where, the place where the actual transmission of data takes places, Whereas, control plane is where the control signals take place. For example, in Netflix, when you go online to put some movies in your queue, that is the control plane, And it carries out itself over the Internet. And the actual DVD in rental services arrives by, say, US Mail service. That's a completely different network, then the Internet, And that is the data point, that's where the actual movie arrives. Okay. So BitTorrent actually uses Internet, P2P, for content distribution, including file sharing, including video sharing. Okay? To share chunks of files and we will come to more details in a minute. In fact, Skype's actual operation remains a commercial secret. There were a few papers trying to reverse engineer Skype operations in the late 2000s. But the BitTorrent was founded by graduate student, and there were quite a few papers written by the founding team about the BitTorrent operation. So we'd know quite a bit more about the BitTorrent and we will say quite a bit more about it in this lecture. Before we go there, I want to highlight that P2P really underlines the tremendous potential of positive network effect. Going all the way back to lecture seven, for example, when we were examining the popularity of a new product and we were looking at the impact of, knowing there are many other users of the same technology and how they influence your decision to acquire the technology or product or service. Now there is the famous Metcalfe's Law. Okay? Named after the inventor of the Ethernet that connects networks, computers in a local area network. And, Metcalfe's law has, quite a few different versions and misinterpretation, But, popular version says that the benefits of joining a network grows like the square of, say n, Which is the number of nodes in the network right now. Okay? So the benefit grows like N^2 order. And the underlying assumption is that every link, okay, are equally important, Which is a problematic assumption, because we know some links are much more important than others. And, furthermore, if I have, five nodes, Pretty much every node talk to every other node. So it forms a full mesh network, And therefore, if you look at the number of connections in the network with n nodes, it is N^2, or, you know, N N - one, which is in the order of N^2. And more generally speaking, we need the number of lengths coming from a node to grow at least linear order of n in order to enjoy this n^2 benefits of Metcalfe's law. And this has been invoked in so many studies in social sciences and as well as engineering. But, do you really need the number of links per node to grow as a linear function of N, not quite. In fact, in P2P law, we basically will see a refinement of this Metcalfe's law. We'll see that even maintaining a number of neighbors that is a constant can work as long as the peering relationships are carefully designed. In the vast material part, we'll see a really careful design through optimization, but even without optimization, you'll see that reasonably smart peering relationship can provide this P2P law. Okay? So, to enjoy the possible network effect, you do not have to have a very dense network of connectivity.
