MOST OF THE MEDIA coverage and blog writings and newsgroup postings and Facebook rantings are made up of ill-informed opinion masquerading as facts, driven by scare-mongering types who fear an invisible big brother looking to crush their Internet experience.
The reality is that most of the information the general public has been consuming on the topics of Internet traffic and network neutrality is, well, hooey.
To try and counterbalance that, we went to someone working in the trenches, as it were, to help explain what is actually going on inside that vast cloud we call the Internet.
Tom Donnelly is co-founder and executive vice-president responsible for worldwide sales, marketing and business development at Waterloo, Ont.-based Sandvine. He has more than 15 years experience in the telecommunications sector and was an integral part of the executive team behind PixStream, one of Canada’s most successful technology startups, which was sold to Cisco Systems in 2000 for more than $500 million.
Sandvine’s technology is sold to cable, telecom, and wireless operators to help protect and improve the quality of experience on the Internet. Its gear mitigates the proliferation of malicious traffic and manages network congestion, constant problems on the Internet.
At last month’s SCTE Cable-Tec Expo, the company released its comprehensive annual report detailing Internet traffic trends where all data included was gathered at the subscriber access network to ensure the findings paint a complete picture of subscriber habits. It’s a report that is huge in breadth and depth, due to the month-long study and tens of thousands of subscribers’ data patterns reviewed.
Donnelly sat down with Cartt.ca editor and publisher Greg O’Brien at the Cable-Tec gathering in Philadelphia. What follows is an edited transcript of their conversation.
Greg O’Brien: Let’s start with your report. What is it telling you? Anything new, different, or surprising?
Tom Donnelly: Anything surprising? Well, we’ve been following very closely broadband traffic trends for the last six or seven years… and the story really has largely been about the explosive growth of peer-to-peer and how it’s really dominated broadband traffic over that period of time.
What’s interesting though is that there was an inflection point early on in the growth of peer to peer when it became more video-based, and that’s where you really started to see it dominating aggregate traffic, upstream traffic, downstream traffic.
GOB: How long ago was that?
TD: Maybe four or five years ago… So, we started with the Napster-like stuff when it’s all about audio file carriage. And then when you have like Kazaa or Bittorrent or stuff like that, increasingly it started to become larger file sizes and those are largely video.
So, I think the story is more about the impact of the growth of video distribution and the increasing demands of higher quality of video content. If we look backwards, that’s predominantly been about downloading of video.
I think what you’re seeing now is this report highlights how higher and higher quality videos are being integrated into wider range applications. So, you’re seeing the category of web browsing and digital media growing faster than peer to peer. Peer to peer continues to be about 44% of aggregate traffic, but if you take the categories of web browsing and streaming, they together represent larger percentage of traffic than peer to peer on its own. And I think that’s interesting.
I think if you look at categorizing P2P and differentiate it from streaming, it’s a little bit complicated – you have categorize these things.
GOB: Yes, because what’s streaming video? Is it the stuff you see on YouTube or is it those video ads that play while you’re reading an online newspaper or something like that?
TD: The real answer is a bit of a mix. The way video works on YouTube is it’s actually a Flash download, but people experience it like streaming. In our report we have actually said you can classify it in two ways. You’d say it’s a popular web site – Facebook and YouTube and large popular sites like that, we’ve classified them as within the web browsing category.
Whether or not they actually deliver their video by streaming or a download methodology is a technical detail that may be less important. And then for the streaming, we’re talking about specifically those applications where things are going right to you
GOB: So, like a Hulu experience or NBC.com where they’re showing their high quality shows with ads around them and all the rest?
TD: Yes. So, the real story is that more and more video is being integrated into more and more applications. Some of it’s streaming, some of it’s downloading, some of it peer to peer. Some of it in web browsing. So, it’s a richer media experience from a subscriber perspective – more and more choices for richer and richer media content, not all of which you have to address simply by downloading content. Some of it you can address in real time.
So, I think there’s a fairly significant increase in the amount of interactive traffic versus downloaded traffic. And more and more that interactive traffic is rich digital media… And consumer expectations of video are increasing, as higher quality monitors, cheaper big screen monitors, for your computer gain popularity.
I guess the underlying story is that broadband is a very diverse multi-application environment. Traffic growth is being primarily driven by the demands for video, and higher and higher quality video. And traffic is increasingly – or has reached the point where it’s potentially – symmetrical. It’s not just coming from some central location. The content is widely distributed. Peer-to-peer makes it easy to distribute it. There’s also now more and more sources of rich media that is coming from the edge of the network.
GOB: Now, what does this mean for ISPs in North America? What does what you found in the report say about how they have to manage their networks differently than they do now?
TD: There are a couple of issues that are common to networks. They are a shared medium, and how you engineer efficiency is you say, “okay I’ve got some resources that I’m going to share dynamically between people. When networks are lightly used, you don’t have to manage them very carefully because you have capacity to accommodate them. But I think what we’re seeing is that application development is at or ahead of network capacity growth. You know, so you have very distributed sources, application development and creativity – people coming up with all sorts of new things in all places all the time and it’s relatively easy to develop them.
But (ISPs) need to insure they manage congestion, that there’s fair use between customers who may have different demands of the network. So, given the richness of applications available, your online experience and what you’re looking from for your service, may be very, very different from other people in your household, other people in your neighborhood, other people on the network. And so that’s a complex issue to manage.
GOB: So what happens next for ISPs?
TD: Where I expect things to go is that increasingly, service providers are going to seek to manage their network for efficiencies, to manage congestion and fairness between customers and deploy tools that will increasingly allow subscribers to set priorities within their own envelope.
GOB: Can you explain that a little more fully? What will customers be able to do when it comes to setting priorities?
TD: So, you would be able to say that for you, VPN is more important than file sharing, so you’re going to do a relative prioritization within your envelope of bandwidth and say “this is my ranking of what’s most important to me. I don’t really care about gaming, so I’m going to rank that lower.”
Or “I live to play online gaming. It’s the most important thing I get out of my broadband experience.” So, companies like Sandvine are investing in technology, that will allow service providers to communicate with their customers and allow them to do some of this self-selecting and customizing their service to their own unique demands, rather than trying to do that just on average across the whole network where you may make 98% of the people happy.
That’s where we are today where I think they make the vast majority of people happy. But they make some people very, very, very unhappy.
GOB: Well, and that’s the issue right now. The so-called net neutrality debate is driven by the 2% that are very, very, unhappy that they can’t get their P2P going faster, or as fast as they had hoped. And to me anyway, it seems that some of these people in a lot of the mainstream media are kind of missing the difference between how P2P traffic acts and how the rest of the traffic on the Internet acts. So, what are you telling people to help make that differentiation and what do you think of the net neutrality debate as a whole?
TD: There are very few neutral parties in the debate… and broadband is a rich multi-application environment. And the network was neither designed, nor can it be managed, to just stay with one kind of user experience because there isn’t only just one kind of user experience.
So it has to make some decisions to balance finite resources. And if you double your network capacity tomorrow, you could spend an unlimited amount of cash over time and just drive capacity, but at any moment in time you have finite capacity. There are just so many different applications that are all seeking relative priorities within that finite capacity. You want to be able to balance, to make your customers happy.
Not all applications are equal in what they demand. The quality of experience is what matters to people… and if you take certain applications like video streaming and VOIP, there’s a point at which one bit less is the difference, not between it’s really good and it’s okay, but the difference between it works and it doesn’t work. So, those kinds of applications require a different kind of treatment.
Of course fast is better than slow, but a little bit slower is… very hard to fall below the threshold where you say, “okay well it doesn’t work anymore.”
So, it seems to make sense in a general sense to say that you know when you have congestion, when you have to allocate resources between impending requirements, that you would first focus on those applications which are tolerant of delays and not focus on those that are intolerant. That’s not to say that VOIP is more important than file sharing, it’s just different.
GOB: Well, a 9-1-1 call is probably more important.
TD: That’s what gets down to. As a network operator you want to create an environment that is optimally efficient to create an environment where there’s the greatest range of choice for people to do what they want. But there will be congestion and when that occurs and where that occurs, you want to have a framework to manage it.
GOB: What is it that Sandvine offers that’s different from say what Bell Canada’s doing or what Rogers is doing? Bell is doing their DPI, deep packet inspection, or whatever they call it but what do you think of that and what does Sandvine offer that’s different that makes network management easier, better, faster, stronger?
TD: I’m not in a position to speak for what individual service providers are doing or not doing but there are a bunch of approaches that you can take to managing traffic. And they have relative advantages and disadvantages.
GOB: (Comcast CTO) Tony Werner talked yesterday about testing some application-agnostic ways to manage the network where everyone is treated kind of equally.
TD: We have a product called Fair Share that helps manage congestion in an application-agnostic way. When there is congestion, you don’t discriminate against—you don’t prioritize on the basis on the kinds of traffic and I think that’s one of the tools you want to use in managing the network, the congestion management. And I call that managing fairness between users.
I still think that there are cases where you don’t want to ignore the characteristics of individual applications. You know, the perfect example of voice and real time video that require different things. You want to be able to maybe take that into account and treat a new 911 call differently. It doesn’t mean you ignore how applications work.
(Customers want to know) is the voice call quality high, acceptable or not? And that’s difficult to do just in managing in an application agnostic way. Which I think, is a long winded way of saying you need to take a belts and suspenders approach to things. You have certain policies where in aggregate you treat traffic uniformly, unless there’s a reason not to.
You can identify what those reasons are and empower your customers to make those choices. It’s obviously a good thing.
But basic traffic management has some form of control, where you know the network’s busy, so you’re going to queue people until resources are available. Or for lack of a better word, shaping, which is when it gets busy, I’m going to reduce the capacity – and there’s relative technical merit to both the functions.
Personally I think that there’s some relative benefit to admission control when you do have access to resources that you get to go as fast as you can. But when it’s busy – it’s like those stoplights on those onramps to highways – where you just waited a bit – and then you get on and you drive at a higher rate of speed.
GOB: The real problem is the growing amount of fear among certain people that Bell or Rogers or Comcast or Verizon are going to be deciding what sort of traffic that moves along in a good rate and moves along at a different rate. To use the highway analogy again, there’s fear that if you’re driving a blue car you’ve got to stay in the slow lane, but a red car gets you in the fast lane. That’s the fear that’s out there. But when you explain to people what the conflict is, it’s extremely difficult.
TD: Service providers manage their network on the basis of the characteristics of the traffic. I don’t really care about the color of the car, who’s driving it or where it’s going. But if someone builds a car that that’s 8 lanes wide…
GOB: Okay, that’s a better analogy.
TD: …then it’s going to be treated perhaps differently. And you know all the analogies to broader networks or you know road networks—
GOB: Power grid analogies, train ones, I’ve heard ‘em all.
TD: They’re imperfect analogies because you’re talking about an IP network where content can be created and distributed and where there are effectively no rules on how you can build the vehicles that run over this network.
Application developers are, and I think rightly so, unconstrained in terms of how they create and develop these services.
GOB: Okay, but a P2P car can be 50 lanes wide or be forced to be half a lane wide, depending on how the network’s managed and what’s running on it right?
TD: Yes. What service providers have been doing, and they get a lot of bad press for this, is they’re trying to balance the requirements of a diverse set of users. But the fact that, you know, you might have been able to go on the highway with your 8 lane wide car a couple years ago and drive at full speed and there’s no traffic on there because no one else was using it doesn’t necessarily mean that you can do that now.
And I think there’s been a lot of confusion in terms of residential broadband is a peak bandwidth service. But I think people think I bought 8 megs, I bought 3 megs, I should get 3 megs, as opposed to I bought access up to that.
GOB: And that’s where usage-based billing is coming in where if you want the guarantees to be at this level and no lower than you have to pay.
TD: Well, if you want guarantees… the only way to really guarantee a minimum is to set the bar low… But the power of broadband and the source of it’s great success has been that you have all of this capacity that can be very dynamically allocated to users at different times and shared.
And like any community event, when things are shared well, it’s good. But if there’s disproportionate sources of congestion – they need to be managed… and, the service providers are the only players in a position to mediate the complementary, contradictory and competitive demands of their many users.
GOB: Are the networks in North America as they stand now, telco or cable, given what’s coming on with high definition video, you know more and more of that rich content that’s going to be delivered to the IP, are these networks ready to withstand and to be able to grow along with these services?
TD: If you look at even the most conservative estimate would say that over the last 7 years, broadband traffic has doubled year in, year out… if you look at it, it’s just grown a lot. And the reality is that it works, right?
By and large, people can continue to do what they want to do. They have more choice. The applications and content is increasingly rich and as much as you might want to complain, it works. And these things grow in parallel. But because it’s not the controlled managed service, it doesn’t mean they necessarily grow at exactly the same rate. There’s always fits and starts, capacity grows, technology evolves, someone comes out with a new application.
It’s a dynamic ecosystem that I think if you objectively look at it, you would say “wow, what a fantastic success.”
In terms of all the different things you could do, by any rational measure, it’s been one of the most successful examples of business growth. But, if you just listen to the recent debate you’d think that it’s about to break, that things are horribly wrong and people have less choice and it doesn’t seem objectively true.
