Internet congestion a reality, Bell says

Canada's largest internet service provider has told the CRTC its traffic-shaping policies are necessary to prevent service disruption for 700,000 customers.

Traffic-shaping necessary to prevent disruption for 700,000 customers

With net neutrality in Canada becoming a hot topic, the debate over how much control service providers should have over internet traffic is becoming more and more emotional.

About 300 angry protesters rallied on Parliament Hill in Ottawa this week to say "hands off our internet" to Canada's large internet service providers (ISPs). Central to the issue is a dispute between Bell Canada and the Canadian Association of Internet Providers, a group of 55 smaller ISPs that rent portions of Bell's network to furnish their own services.

Bell in March extended its traffic-shaping, which involves the slowing down of certain peer-to-peer internet applications at peak times of the day, to the CAIP members, prompting the group to complain to the Canadian Radio-television and Telecommunications Commission. The CRTC declined to issue an emergency order to stop Bell from using its management technology, but did open the issue to a public inquiry.

Bell says it needs to manage its network in order to cope with ever-increasing traffic, while the CAIP says the company has failed to prove there is a congestion problem necessitating such measures.

Bell on Thursday night filed documents with the CRTC that spell out the congestion issue. Mirko Bibic, head of regulatory affairs for Bell, discussed the submission with on Friday.

Mirko Bibic, head of regulatory affairs for Bell Canada, says the protests against large internet service providers have been fuelled by misinformation from people who don't run networks. ((Courtesy of Bell Canada)) CAIP says by their measure, which are latency and dropped packets, they haven't really detected any sort of congestion on the network they share with Bell. But according to your filing, you're mainly using link utilization and by that measure there is some congestion. Can you explain what the difference is between those two measures?

Bibic: A link is the fibre facility between two nodes and the network. Nodes are generally switches or routers. There are a number of links in our network and, as you pointed out, our network is shared with the wholesale ISP customers of ours who choose to subscribe to our Gateway Access Service (GAS). So they're linked to various points in the network – there are DSLAM links, aggregation links, broadband access server links and backbone links. There are all those types of links. In fact, our answer that we filed with the commission explains that in more detail.

There is a correlation between measuring the level of utilization of a link and the issue of latency and dropped packets, because as the level of utilization of the link increases, the number of congestion events become longer in duration, and that ultimately will lead to latency and dropped packets. There is a correlation there, but what we've done and always do (and this is accepted or common practice in the industry) is recognize that we have links at several components or areas in our network. What we do is we want to make sure that as the utilization levels of the links approach a particular threshold, which is less than 100 per cent because you don't have to get to 100 per cent before you start experiencing a congestion event.

As it gets up to that magic threshold, which we've also filed with the CRTC, we've seen these latency packet events. What we've done therefore is engage … in network management to relieve that congestion during peak periods.

You'll also note that in one of our answers to the CRTC that was filed last night, there is a table that we've provided and the table explains packet loss events or how that's increased over the last number of years as the result of increased utilization or increased traffic on our network. Of course, our primary focus is that the majority of users get a robust internet experience, and network management in this manner allows us to do that. You mentioned a chart – is that one of the charts that has been filed in confidence with the CRTC?

Bibic: No, that chart is in the public domain. We filed a public version of that chart in the answer to the second CRTC question. CRTC 2, it's called. One other thing that wasn't clear from this filing is that there was an estimate that 700,000 customers will experience some sort of slowdown this year unless some measures were taken. Can you somehow better quantify this congestion problem?

Bibic: If no measures were taken, then 700,000 customers would have been affected by congestions during peak periods. We want to obviously take steps to make sure that doesn't happen. So this network management is, as we've stated, one of the ways to address the issue of congestion during peak periods.

I do want to emphasize that this phenomenon is occurring primarily during the peak periods.

Another thing we're doing … is continuing to invest in our network, and we invest more than any other provider. Those are two of the methods to make sure customers aren't affected. Our filing to the CRTC gives data on the typical number of customers that could be affected by congestion at only one link. And there are different types of links, like I said, and certain links have different capacity levels. There are high-capacity links and lower-capacity links. And the higher-capacity links, obviously if they get congested, would affect more customers. So getting back to your question, the CRTC has some more precise figures on the number of customers that could be affected by type of link if congestion during peak periods is not addressed. One of the complaints in regards to that is that some of this [traffic management] is happening without a lot of transparency. Obviously a lot of these figures have been held confidential with the CRTC for competitive reasons, but what's your feeling toward this concern?

Bibic: I've stated before quite clearly, as has my colleague who runs the wholesale business unit, that when we started applying the network management practice to the end users of our wholesale ISPs, when that happened and resulted in the CAIP complaint, my colleague wrote a letter to his customers regretting the fact that we hadn't advised them in advance of taking this action. Since then, we've opened a dialogue with our wholesale ISP customers to address their issues and we're quite open to continuing that dialogue.

At the end of the day, the wholesale ISPs are our customers and we generate revenue [from them], so we want to make sure we're serving them to the best of our ability as well. We've also indicated in yesterday's filing with the CRTC that we are prepared to inform impacted wholesale ISP customers who subscribe to our GAS service – at the latest on the day of implementation – if we plan to make material changes to our network traffic management practices. So clearly transparency is an important issue, we're aware of it, we're sensitive to it and we're going to take steps to make sure that becomes a primary concern going forward at all times. Seeing protests on Parliament Hill against abortion or war is not uncommon, but seeing one about internet access is. Do the people who are up in arms over traffic shaping by Bell and Rogers and a few others have it wrong? Why are they protesting?

Bibic: There's been a lot of misinformation about what Bell has actually been doing and why, fuelled in large part by people who don't operate in the real business world and who don't have to manage multiple factors including constructing networks, operating networks, managing networks, responding to consumer desires and offering consumers services they want, when they want, at the appropriate pricing and responding to competition. There are a number of factors that we have to take into account and manage in the real operating world and there's been frankly a lot of misinformation about what we're doing.

As an example, quite frankly, the speculation about what deep-packet-inspection technology allows us to do with respect to inspection of packets and whether or not that constitutes breach of privacy. We've been quite clear all along that we don't examine the contents of what the packets contain, yet there are these applications being filed with the Privacy Commissioner alleging that we do. That's just an example. We actually welcome the opportunity the CRTC has given us to add facts to the record and the public discourse on this issue.

When the CRTC started the formal process related to the CAIP application after the interim application was denied, I stated publicly that we welcomed the opportunity the process was giving us to put the facts on the record, and we've started doing that. Yesterday was one such opportunity. If you take a step back, I think it's important to look at what we said all along and what the evidence shows. We stated all along that there is a congestion issue during peak periods and the evidence we filed yesterday shows that. We show where there is congestion, how we measure it and we explained how deep packet inspection works.

We stated all along that network management was one of a multi-pronged approach to addressing ever-increasing bandwidth consumption and we've put on the record cold, hard facts to the extent of our investments in our network and the correlation between our investment and traffic growth and the consequent impacts.

I think at the end of the day, the CRTC is going to rule based on the evidence based on the record and the evidence is compelling that what we've done is a reasonable and appropriate way of managing our network. Last week you introduced a video store, where customers can download movies. Many are wondering how can you say you have network congestion problems and throttle peer-to-peer applications, then introduce a service that is going to add to that problem?

Bibic: The Bell video service, the content that we will be distributing, isn't using peer to peer. Peer to peer is by design a high-bandwidth use application. What we're using here is content-distribution network technology and we think you'll find others, as they enter this space, will likely use this technology as well.

At the end of the day, certain people have been wondering whether or not network management was implemented to support this particular Bell video service and that's just categorically not true. First, let's look at the market in which the service will operate. We're operating in the video distribution market that comprises bricks-and-mortar video rental stores, DVD sales, pay-per-view movies on cable networks and Bell ExpressVu. The host of competitive options for consumers to watch movies is numerous. This is just a nascent type of service. Others are going to jump into this kind of distribution method. Good, that's just going to enhance competition and customer choice, but certainly network management practices are in our view unrelated to this and are not going to have an impact on the ability of others to enter into this kind of downloadable movie space and compete. Obviously this is different technology than peer-to-peer, and competing services such as iTunes or Xbox Live aren't being throttled, but at the end of the day aren't the different technologies still filling up the same pipes, which Bell says are congested?

Bibic: That goes back to the time-sensitive nature of the traffic. We've discussed before the reasons why we've decided to focus on peer to peer, which by design is a high-bandwidth use application, and also less time-sensitive. If you're going to be downloading movies via peer to peer, by design – and I'm no technologist – you're not talking about viewing that movie in real time. Just by the nature in which the movie is going to be downloaded and how it's going to get to them. Nothing is preventing someone from downloading a movie with peer to peer and nothing is preventing them from downloading it during peak periods, it's just going to be a slower download. It'll still be downloadable. The other option is to download it outside of the peak periods. There are a number of options.

CBCNews: One of the questions raised by a recent OECD broadband report is why isn't Canada spending on fibre to the home, which many feel would alleviate much of these bandwidth issues. In the U.S., of course, you have Verizon doing so. Why has Bell opted for the cheaper fibre-to-the-node option, especially if there are congestion problems?

Bibic: The notion that investment in the network, whether it's fibre to the home or fibre to the node, is going to solve the problem… we've got to keep in mind, and there is evidence now on the record of the hundreds of millions of dollars that we spend on our network, over $3.2 billion in the last few years on DSL infrastructure alone, and a lot of that is on fibre to the node. So we're already spending hundreds of millions of dollars. We're spending hundreds of millions of dollars in upgrading and ensuring that our ATM network remains robust, and that's our legacy network.

You've got to keep in mind that in an ideal world, you want to direct most of your capital dollars to next-generation technology and not the legacy network, yet we're doing it to ensure that the majority of our customers who are currently on the ATM network continue to have a robust service. There is also a timeline between building and making the new network operational, so while you're building you've got to continue maintaining the old network.

You have to appreciate that capacity will always be used up. We live in a dynamic world, not a static world, so we could increase the capacity of our pipes, which we do more than anyone else in terms of investment, but we can't predict what new application is going to come along, what capacity that's going to consume, and the ever-growing desire by consumers to consume more and more and have a rich internet experience, and that's good. It's good for consumers, I have no quarrel with that. The point is, it's not a dynamic world so we'll build the pipe, it will get used. Building alone is not going to solve the problem.

We'll have to be realistic here and the answer lies in building, in managing the network, in pricing plans as well, and it's not unlike congestion on a highway. If you have a two-lane highway and you have congestion at rush hour, you're not going to build 20 lanes because those 18 other lanes just won't be needed during non-rush periods. So what do you do? You build a couple of extra lanes for one, you expand the infrastructure. As well, you do things like have bus lanes that allow buses, taxis and cars with more than three passengers to travel on them so that they get faster service than if you choose to drive your Escalade and you're alone on the highway. You get to go on the highway and you ultimately get to your destination, you might just be a little slower. There are a number of issues here. As far as fibre to the node and fibre to the home, I think it's best for me to leave that to somebody else within the company.