Day six of the CRTC's network management hearings opened with a final consumer group (Union des Consommateurs) and closed with three of Canada's biggest ISPs – Rogers, Videotron, and Shaw. Bell was scheduled to appear today but has been pushed back until Tuesday.
The big storyline of the day was the disclosure by Rogers and Shaw of previously undisclosed information. Rogers revealed its traffic management practices (throttling P2P upload speeds) and shockingly admitted that all its tiers receive the same upload treatment, regardless of the price paid by the consumer. This is true even though its promotional material tell customers that higher tiered service offer faster upload speeds. Shaw disclosed that it engages in similar practices and provided insight into its throttling practices, noting that it guarantees 80 kilobits per second for throttled P2P sessions and that it reserves 30 percent of its bandwidth for P2P use (it said that 10 percent of its users account for the P2P traffic).
Videotron, the third cable ISP in the mix, complicated the analysis further by noting that it does not traffic shape. Rather, it uses economic measures, the new euphamism for bit caps, to discourage overuse of P2P. The ISP indicated that it is very happy with the effectiveness of its approach.
Today's summary was again compiled by Sean Murtha, a law student at the University of Ottawa. Other coverage available from the National Post liveblog, CBC.ca, the National Post, Cartt.ca, and twitter feeds from CIPPIC and me.
CRTC Net Neutrality Hearings – Day 6
Union des Consommateurs
The Union des Consommateurs (UdC), in their opening remarks, suggested that there has not been enough evidence of actual congestion, which many ISPs have complained of this past week. The UdC said they wanted the CRTC to consider many issues, including concentration of ownership, quality of service affordability, and competition. UdC also said that one of the real reasons that ISPs are targeting P2P is because the existence of P2P, and its use by customers, threatens other businesses. They argue that the whole issue of traffic management is about competition, not congestion. Thus, discrimination as a result of traffic management will have the detrimental effect of stifling innovation.
UdC said that to motivate network management method of throttling a specific application, there should be clear measurement-based evidence that:
- there is congestion in the network;
- time sensitive applications are experiencing unsatisfactory Quality of Service; and
- the throttled application is a significant cause of the congestion.
There is little empirical evidence that these conditions exist. Moreover, UdC argued that it is not clear from the presented data that P2P applications are the primary cause of the congestion, if it does exist.
The UdC stated that new data on recent traffic volumes does not show the exponential growth that was predicted in 2006 and 2007. The rate of growth is decreasing, and P2P traffic volume is growing at an even smaller rate. The UdC argued that “modest growth rates” in usage mean that increasing network traffic can be accommodated by reasonable increases in network capacity. If it is determined that network management is necessary, there are several options that do not involve "application-specific throttling". For example, during periods of congestion, it is feasible to control the rates of individual users, reducing the rates of high-volume flows irrespective of the protocol employed.
UdC also discussed deep packed inspection (DPI) vs. shallow packet inspection (SPI). The method used by most ISPs is DPI, which involves processing the payload and the "application layer header", searching for known signatures associated with specific applications. UdC argues that this practice raises privacy concerns, threatens robustness by violating layer design principles, and promotes the practice of data encryption.
UdC explained that SPI does not involve any examination of packet payloads. Instead, statistical features are collected for each flow and, based on these features, the flows can be categorized into application classes. Experimental studies have shown that this classification approach can be very accurate, but the classes must be more general. UdC stated that the classes are sufficiently specific for network management. Also, UdC said that the technique alleviates any privacy concerns.
UdC concluded by saying that ISPs can adopt positive approaches to improve the way P2P applications operate, benefiting both the users and the access network. The P4P project promotes cooperation between P2P applications and the ISPs. Field tests indicate that the methods both improve the performance perceived by the user and reduce cross-ISP traffic and the load on ISP aggregation links.
Chairman: You mention that there should be clear evidence of congestion. Is this data publicly available? Is the data difficult, or economically unfeasible to collect?
UdC: The evidence is certainly collectable. It is more expensive to collect than utilization data, but it is certainly able to be collected. Whether the companies collect
Chairman: Is shallow-packet inspection being practiced by anybody? It seems that everybody uses deep packet inspection. Is this being employed?
UdC: As far as I know, nobody uses SPI exclusively, but some ISPs do use it along side DPI.
Denton: Various ISPs have said that if there is going to be network control, it should be exercised by retailer, not the wholesaler. What do you say to this?
UdC: We question the motives of the ISPs. If the issue is congestion, that can happen at the retail level.
Denton: Does traffic management interfere with innovation?
UdC: By allowing P2P, which is often targeted by traffic management, we can help with innovation, as P2P is used for very innovative purposes. Targeting P2P can hinder innovation.
Denton: What is your final, take away message?
UdC: There needs to be clear measurement studies to show what is happening inside of the networks. We need to find if there is congestion, the root causes of congestion. SPI is an intriguing thing that we should look into.
Lamarre: In regards to SPI, is this the first step in DPI? Should ISPs simply stick with this level of inspection (i.e. SPI)?
UdC: SPI is actually the 2nd step in DPI. My suggestion would be, based on studies, SPI is very effective. Get over 95% accuracy of classification of flows, without looking inside of the packets.
UdC: We have to remember that we must look at the internet as a public service, a public good (or a public utility).
Rogers Communications Inc.
For Rogers, Ken Englehart, Senior VP Regulatory, appearing with others. Rogers noted that Canada has the highest penetration of cable modem service in the world and the highest penetration of broadband in the G8 countries. All of this has been accomplished with a minimum of government regulation. Rogers argued that market forces, not government fiat, are responsible for this. Rogers pointed out that many participants before the CRTC "want lawyers, not engineers, to design the networks".
Rogers stated that they are committed to an open internet. They stated that they do not manage their networks to favour their cable television business, demand fees from internet services for access by their subscribers, or stop independent documentary filmmakers from distributing their content. They do not open email, inspect the contents of packets, or focus on specific packets. Rogers would urge the commission not to set internet traffic management guidelines at this time.
Rogers noted that the FCC rejected this approach (i.e. setting guidelines) in the Comcast decision. As Rogers stated "the internet is to new and is changing too quickly to establish ITMP guidelines at this time. It is better to adopt the FCC’s approach of looking at individual internet traffic management practices on a case-by-case basis. In that way, the Commission will ensure that Canadian ISPs have the flexibility they need to manage their networks properly." Rogers pointed out that their traffic management practices do more than simply "traffic shaping". Rogers does much to protect customers’ experience. Rogers says they protect the network from spam, prevent denial of service attacks and virus attacks, and blocks access to child porn sites.
Rogers stated that they do rate shape upstream P2P. They said that P2P can "clog the highway", and the traffic needs to be managed, to make it fair for other users. They pointed out that it was not simply because it was P2P that they managed it, but because "this protocol is designed to swamp the network, and this is why it must be managed". Rogers said that their practices do not simply favour their own services, and argued that without their traffic management practices, competitive VOIP providers would be hurt by congestion. While some interveners have claimed that Rogers should simply increase capacity, this would mean they would have to provide a huge amount of additional upstream capacity. Rogers argues that this would significantly increase their costs, which would then increase the cost to all consumers, not only those who use P2P. Rogers argued that they "believe that it is readily apparent that our traffic management practices are an appropriate network management solution".
Rogers also argued that that "have found that…customers have very little curiosity about the inner workings" of their traffic management practices. Customers simply want fast reliable service, and do not inquire any further than that. For wholesale customers, Rogers stated that they believe that there should be sufficient disclosure to allow them to understand the nature of the services they are acquiring and to respond to their own customers’ inquiries appropriately.
In regards to privacy, Rogers stated that they are concerned with privacy, and work to protect them. They argued that privacy issues raised by consumer groups concern things which ISPs could do or might be able to do, not what they are actually doing. Rogers urged the CRTC to look at what ISPs are doing.
In regards to wholesale services, Rogers noted that their wholesale services are not currently traffic shaped, but they reserved the right to do so in the future. If wholesale traffic begins to impact their customer’s traffic, they will revisit this decision.
Rogers noted that they do not currently rate shape their mobile wireless data traffic, but think that with the growth of use of their wireless services, traffic management will be necessary in the future. However, because costs are different, architecture is different, etc., it is too early to identify the traffic management practices that will be required for mobile wireless carriers.
Rogers also addressed section 36 of the Telecommunications Act. Rogers’ traffic management practices do not control the content of P2P file sharing traffic. While it may take longer to download a file, this does not control the content or change the meaning of or purpose of the file.
Chairman: What the CRTC is talking about is an analytical framework. Would it not be good to lay out a clear analytical framework, would this not be helpful for the industry as a whole, including Rogers?
Rogers: Without guidelines, we have gotten along just fine. Where there were guidelines, they came about after decades of experience. Secondly, some guidelines put forward by interveners, like the Oakes test, have a bias against network management. This is dangerous. If there is going to be an analytical framework, it should focus on the customer, it should focus on the customer experience. Also, we should not assume that carriers are trying to upset customers; there should be a presumption that network management are for the benefit of customers. Moreover, look at what other ISPs are doing. Other ISPs are doing the same.
Chairman: If Rogers thinks that the analytical framework as put forward by interveners are not ISP friendly, then Rogers should suggest a different analytical framework to the CRTC.
Chairman: The chairman wanted clarification on technical issues, such as SPI and DPI.
Rogers: There was a differentiation between DPI and SPI. The way that UdC described it, the first level was deterministic, and then the 2nd stage (SPI) was the classification stage. Rogers uses a combination of both techniques (DPI and SPI).
Chairman: But the studies we have been told about show that SPI is effective, and accurate.
Rogers: We have not seen the rates of accuracy discussed this morning in regards to SPI (95%). Rogers has seen much lower accuracy levels using DPI.
Rogers: P4P is a working group. The challenge with these working groups which tries to come up with "network aware" ways of dealing with P2P, it relies on users. For users who do not rely on guidelines, they continue to use P2P the same way they did before the guidelines were in place. It relies on everyone being a good and responsible actor, and that is not the case on the internet.
Chairman: Like Cogeco, you said you only shape upstream traffic. Is it mostly your own customers who download? By shaping upstream, do you also influence the downloading of customers?
Rogers: Obviously it will happen sometimes. However, Rogers said that they tested this issue in a lab after the question was asked to Cogeco on Friday. Rogers used a computer connected to the internet with a single video file. 20 worldwide users were downloading the video, but none were Rogers' customers, and none were Canadian users (users from U.S., Ireland, and Finland). So, if Rogers traffic shaping was so oppressive, why would so many users be accessing the computer?
Molnar: Rogers spoke about guidelines, and whether there should be any. One of the issues as to not having guidelines is the fact that if there is discriminatory practices, particularly as it relates to traffic management that is specific against content, if such practices exist, by the time an ex post review happened, you could cause a major economic impact on the targeted application.
Rogers: There is a danger to setting up guidelines. The internet it so new, applications are arriving so quickly – there is so much going on, guidelines being set up now could quickly become obsolete, and would hinder ISPs ability to help their customers. While there is fear, the actual evidence shows that ISPs, to be competitive, have to address customers’ needs and concerns. Net neutrality concerns have been overblown.
Molnar: The OIC felt that discrimination against an application is discrimination under s. 27(2).
Rogers: Rogers does not think that 27(2) is supposed to protect applications themselves, but is meant to protect application providers.
Molnar: Rogers throttles on a 24/7 basis. So, help me understand why it makes sense to throttle on a 24/7 basis, instead of when the need exists?
Rogers: It is useful to think of the analogy of the network as a highway, and traffic on the network as cars. People may say that if there are too many cars, you should build a bigger highway. In Rogers view, this misstates the problem. Rogers wants more traffic, and they continue to build new capacity, although it is expensive. The problem is if one car parks on the highway, or reserves an entire lane of the highway for themselves.
Molnar: Using your analogy, you have built a super highway, and if one car is parked on a lane for 1 hour, why do you feel the need to throttle the network 24/7?
Rogers: The problem is the behaviour of someone who abuses the system constantly. Congestion problem we experience is at the node level. But the traffic management happens at the router, which covers many nodes. Congestion may happen at any node at any time of the day, and that is why we practice 24/7 throttling.While on average the network has "busy periods", a small cluster of users can congest the network at any time of the day. We have tried to build our way out of this by increasing capacity, but P2P users have continued to absorb capacity within the network. We have done tests where we turn off P2P traffic management in certain clusters, and even at 3am congestion has occurred in these test areas. While we traditionally think of peak utilization in the network, we actually do not have peak points of time, but peak periods of time, which can last for hours.
Molnar: Molnar wanted Rogers' data on P2P traffic. She asked them to provide the CRTC with their data.
Molnar: What is the rational for throttling only on the upstream, and not on the downstream?
Rogers: We think we can control downstream flow through augmentation and pricing. On the upstream, because we are constrained by a cable modem network, it cannot be controlled. On the upstream, it is the "rest of the world", not only our customers, using the network.
Molnar: Rogers commented that the Comcast agnostic approach would be more expensive, and would be more troublesome than the targeted approach which simply targets P2P. Can you elaborate?
Rogers: The Comcast methodology restricts all applications. If we used this, we would have to limit what our customers could do. It would be more costly than it would be today.
Molnar: Exactly how do you throttle P2P?
Rogers: It is the same for all packages, the same bandwidth allocation (the same speed) for all packages.
Molnar: I want to talk about the issue of disclosure. As I understand, you have only recently begun to disclose to customers throttling practices. You say that Roger’s customers "have very little curiosity" about traffic management practices. Do you not think customers may want to know more about throttling practices, especially if they want to P2P? There is nothing in the package descriptions on the Roger’s website which mentions throttling, even in packages which are targeted to people who want to download a lot, and play games. If you are concerned about the customer, shouldn’t they be informed?
Rogers: Our different services have different cap sizes. What our descriptions are saying is that people who are concerned with downloading and game playing will want a package with a higher cap. We feel that our throttling information is still competitively sensitive information. If the CRTC feels strongly about it, we could reveal more information.
Molnar: Some other ISPs have said that competition should address ITMPs, and this could actually be a form of differentiation between ISPs.
Rogers: There is a very competitive market in Canada, and all of the carriers are making huge investments to stay competitive. Surely traffic management is part of this competition.
Molnar: What about traffic management as it relates to content management and protocol? Will this become part of the competitive marketplace?
Rogers: We are measured by subscribers based on the performance of the network, and users want a wide range of things which they can do using the internet. Rogers benefits greatly from an open internet that continues to evolve.
Molnar: You mentioned that you do not rate shape on your network. Are there differences between ITMPs on wireline and wireless?
Rogers: Because behaviour of users is different the solutions may be different, but the options would be the same.
Molnar: What is it when we talk about providing preferential content, or content under a different pay structure?
Rogers: All ISPs, wireline or wireless, have a homepage with content on that homepage. Rogers has a homepage, with certain partners on the homepage, who work with Rogers. The technology has evolved quite a lot, to the point where customers can go to any page, not simply the homepage of the provider.
Molnar: Is there any traffic management associated with this activity?
Rogers: No, there is not.
Katz: While you said the FCC has done nothing at all in regards to this. What did Comcast do? What would the cost be if you were to follow the net neutrality statement released by the FCC in 2005 which caused Comcast to change?
Rogers: The FCC statement dealt with net neutrality, not traffic management. Comcast denied that they were engaging in packet reset methodology, and this did not help. They hired people to come and clap at the hearing, and this behaviour annoyed the FCC. The FCC did not tell Comcast what to do, and most US companies manage their traffic just like Rogers does. Rogers does not think the Comcast methodology is good at all. We think it is a mistake. In terms of the cost, it could be "fairly huge".
Katz: What would it take if you were required to follow the Comcast methodology? Could you explain what Comcast was doing, what did they change, and what did it cost them?
Rogers: Comcast was doing "session management", sending out phony "reset packages" saying that sessions were no longer needed, when in fact they were. According to evidence at the FCC, this really only was meant to slow things down, not stop the flow completely. What Comcast is now doing is using the "switch" that cable modems do. It looks at congestion at the individual node, and will slow down individual users who are nearing their capacity for 15 minutes. It targets P2P users, but slows down all of their traffic, not only their P2P. Rogers does not want to do this, because it would be bad for customers, because not only would they be able to use P2P, they may not be able to make a VOIP call. As to what it cost, Rogers does not have that information, but they will try to get it from Comcast for the CRTC.
Katz: Or does this tariff limit the ability for wholesale customers to sell value added services?
Rogers: As long as the resellers are ISPs selling IS, they can enhance it any way they want.
Katz: Rogers has a usage based wholesale tariff for ISPs. When you are charging based on usage, why would you use traffic management, when they are paying for what they are getting?
Rogers: We do not traffic manage them, but we reserve the right to do so. Their impact on the network is the same as one of our customers.
Katz: Rogers reserves the right to manage, monitor content posted. Does this mean that Rogers will go beyond DPI, and actually monitor the content?
Rogers: Rogers does not do that right now.
Lamarre: The way that Rogers is describing P2P leads me to great reservations. You gave the example, several times, of downloading movies, as if this is the only thing that goes on. Also, you discussed the experiment. However, this is nothing more than "anecdotal information", this is not scientific data. Do you have empirical, scientific data that most users doing P2P are doing it 24/7, and whether this goes beyond the actual use of the bandwidth sold to the customer?
Rogers: It cannot go beyond the bandwidth sold to the customer. Rogers said they would try to get the CRTC more scientific data.
Lamarre: In regards to wireless, you mentioned you are not doing traffic management for wireless the same way that you are doing on the wirelines.
Rogers: You could say we are using price to manage it.
Lamarre: Would you say that SPI would have less impact on privacy than DPI?
Rogers: As mentioned, we are using SPI, as well as DPI. There seems to be a bit of a misunderstanding in regards to DPI. DPI does not keep any records of which customers were doing P2P, it does not look at content, and it simply classifies the data. Some DPI methods do raise major privacy concerns, and we could do that, but we do not. It is against the law, and there is no business reason to do it.
Lamarre: But you do monitor. To sort, you have to monitor.
Lamarre: Do you keep aggregated data as to what is going on in the network?
Rogers: We have to keep aggregated data to charge customers who go beyond the cap.
Denton: Different narratives exist. The CIPPIC narrative states that traffic congestion can be addressed through increases to capacity and augmentation. Also, some experts have argued that we should not overact to temporary problems. You have said that rational people will fix the problem, and this has led to throttling. Why does CIPPIC, and the OIC, not get the story right?
Rogers: We are always working to augment growth, and are spending 10s of millions of dollars a year in doing so. We are trying to stay ahead of congestion. Secondly, we understand why some are concerned about application specific traffic management. However, this is not a network neutrality issue. We have a network which is designed to be a shared use network, and the fact is that we have to manage it. When a certain application tries to use it too much, users who use those applications cannot reasonable expect that to be ok. It is the behaviour of the application, not the application itself that we are concerned with. If an application which could cure cancer acted in a certain way, it would be also be subject to traffic management. Rogers would agree that we should not overreact to temporary problems. If we are only trying to help our video business, why would we not be targeting the downstream? We disagree with the narrative as it has been presented my many interveners.
Chairman: Are there other ways, instead of throttling, to control the upstream?
Rogers: Before traffic shaping, we were in a state of congestion. Pricing mechanisms would not work as well as what we are doing now.
Chairman: Why would a pricing mechanism work on upstream? Is there no economic solution, instead of traffic shaping?
Rogers: You would have to have a very low cap for upstream, and you would have to educate your customers about the true nature of P2P (i.e. that other users access their computer for information). Also, you would need a sophisticated pricing mechanism. Moreover, Rogers is not aware of other companies which have acted in the way that you propose.
Rogers: An unmanaged network is not a neutral network. It does not make sense to say that because you are engaged in application traffic management that you are not neutral. There are always decisions that we make that which favour some users over others (i.e. latency decisions which affect gamers), but this is clearly not traffic shaping. We are happy to explain to the CRTC, in private, about our traffic shaping measures.
Quebecor Media, on behalf of Videotron Itee
Videotron (VT), the largest service provider in Quebec, explained that they have invested over $100 million in their networks. They said that if only 1 thing could be taken from their presentation, it should be that the internet is a dynamic and evolving medium, and that the CRTC should not impose regulation. Also, they stated that they do not currently use technical ITMPs, and to this point they are happy with the use of "economic measures" to deal with potential problems. However, VT refused to rule out the use of technical measures in the future.
VT made it clear that regulation should only be implemented if a "real problem" is identified. They argue that, at this time, there is a lack of proof as to actual problems. The "problems" which have been discussed up to this point are, according to VT, only "hypotheticals", and such hypotheticals do not justify regulation.
They noted that resellers that have participated in the CRTC proceedings have generally maintained that ITMPs, both economic and technical, are wholly legitimate. In their view, underlying networks should apply these practices only to their own end-users and not to their resellers end-users. VT called this a "rather self-serving position", and pointed out that on a shared network, when one end-user or group of end-users causes congestion on a particular facility, the impact is felt by all end users who use the facility. It is the network operator, and not the reseller, who must support the high costs of physically splitting the cable cell to increase capacity. VT argued that a network operator which is refused the ability to apply its practices symmetrically would soon find itself the target of resellers in search of arbitrage opportunities, and these resellers would have an interest in attracting those end-users who seek to circumvent the network operator’s traffic management practices. VT explained that the excessive cost arrangement would be supported solely by the underlying network operator, and all other end users would suffer degradation in service. Such a result cannot, VT asserted, cannot be in the public interest. Also, there is a healthy competition in ITMPs. This underlying competition is sufficient. An intervention by the CRTC cannot be justified because of a lack of proof as to actual problems.
Chairman: You have mentioned that you use financial means to control use, and you have thresholds, and excess use is charged as surplus. Rogers said that these surplus charges do work to control downloading, but the problem is with uploading, and there is no financial means available to control uploading patterns. That is why they rely on technical means. What has your experience been? Do you act as Rogers does?
VT: We do not use technical means, such as DPI, like Rogers does. Indeed, you are asking us how efficient our financial tools are. It is true that, theoretically speaking, using a usage threshold over the course of a month should not impact users. In fact, we do see that there is a correlation. The implementing of an uploading threshold does have an effect on overall usage. For example, it changes the amount of information which is uploaded. It is a strong enough correlation to show us that this is effective.
Chairman: Rogers has said if they do not use traffic management, P2P will flood their networks causing congestion. Do you educate your users, do you tell them that you can block uploading and can restrain the amount of uploading?
VT: No, we do not have a plan to educate our clients. Over the last few years, since 1999, we have been using usage thresholds, and our clients have been very happy. Videotron clients are informed, they know how the internet works, and they know that uploading is more limited than downloading. Clients know that P2P can clog networks. Users either limit the amount they use P2P, or they disable this feature.
Chairman: Do you have precise data on uploading and downloading, because you are telling us the opposite that other carriers have told the CRTC.
VT: We will see what data we can supply. We have some data. When we have instituted thresholds in the past, I am sure that we have tracked some data. However, our comments are not in contradiction of other suppliers comments. They have moved towards technical traffic management methods. But the implementation of thresholds has kept congestion under control, but we are not saying this is the only method.
Lamarre: In regards to economic means to shaping traffic, whenever we have raised this issue, we were told about 2 problems. The 1st problem was to publicly educate consumers, the 2nd was to quantify customer satisfaction as to economic means of management. You have said that 98% of users as per Léger marketing said they were happy. Could you give this survey to the CRTC?
Also, how does the usage meter work?
VT: All the user has to do is identify themselves, and they can access the meter and get user data from it. We were among the first to track usage, and this usage tracking system is fully functional.
Lamarre: You have 1 million clients, do you have 1 million meters?
Lamarre: If the users have a complaint
VT: The main goal of these meters is to allow customers to monitor their usage. If there is a complaint, then it can be investigated. It should be noted that these meters are made available to resellers.
Lamarre: How much does it cost for these meters, per user?
VT: I am sure we have that data.
Lamarre: What is the life span of this equipment?
VT: We will try to get you that data and information.
Lamarre: You have said that Videotron reserves the right to change its rates when it is part of a new offer that has not yet been communicated to the public before it comes into effect. What do you mean by this?
VT: This is in case Videotron has a new product, with a new usage threshold, this is the type of offer we do not want to announce ahead of time so that our competitors get a sneak preview. We have inserted this clause to protect ourselves.
Lamarre: What is the link between this and your traffic management practices?
VT: If management practices include economic means and tools, then you understand that we cannot divulge economic tools ahead of time, because that would have a detrimental economic effect on us.
Lamarre: In your response, you said that Videotron has 6 packages. Despite the fact that you do not impose maximum bandwidth quotas, you do not have a one size fits all system.
VT: Our clients do have a choice of 6 packages, and this benefits the consumer. This choice is entirely to the client’s benefit. A user that needs to download more than upload would choose a package which suits their needs best, while my Grandmother would choose a different package.
Lamarre: In your submission, you say that it would be unreasonable in terms of relations with the customer to require an extensive disclosure of methods used. What do you anticipate would be non-productive, in relation to relationship with customer.
VT: Our main concern is with the complexity of the information. It would force us to post very complex information, and would be very confusing, and would lead to calls to our customer service centre. A certain disclosure yes, but not to that extent.
Lamarre: So you are trying to strike a balance, without having to give consumers a course on network engineering?
Lamarre: Even if you do not have technical measures in place, you do have economic management techniques. Your position is to promote full openness of the internet, which is normal. To defend this position, you say that there is an unpredictable side to all of this. What kinds of predictions do you make to try to handle your internet traffic? Are you thinking about offering different kinds of bandwidth elements? If we are trying to make predictions as a good ISP, what methods are you using when you decide whether or not to expand your network?
VT: Any network manager will try to predict the total load on their network. The way we do it is to provide for the needs of the network 6 months in advance, just like Rogers. We do not rely uniquely on our economic management techniques. There is a lot of unpredictability, we cannot be sure of customer behaviour a year from now. Our main message is we need flexibility, we need the ability to change quickly.
Lamarre: While you say it is unpredictable, Rogers says it evolves very quickly. If we do not do anything, the Canadian population and your customers may be faced with ISP which cannot face the demand.
VT: It is in interest of ISPs to offer sufficient services to customers, in this very competitive Canadian market. We do not see that market forces are such that ISPs have an incentive to degrade services to the customer. It is possible companies will make mistakes, but these mistakes will be fixed. If mistakes are made by a certain company, but their competitor does not make a mistake, then customers will shift to the company that does not make a mistake.
Lamarre: You mentioned quickly in your presentation about services you offer to your wholesale clients, the impact of bandwidth has been discussed at length. Do you enforce your economic management practice
VT: Yes, in our agreement with our resellers, there are surcharges (TPIA tariff) for people who go over their maximum. We do not enforce this on the end-user, but on the wholesale customer.
Lamarre: Currently you encourage your reseller to implement the same system of billing according to bandwidth, correct?
VT: They have access to the counter mentioned earlier, but it is up to them whether they use it.
Lamarre: The bandwidth offered to the reseller is not isolated?
VT: It is the local access where the congestion will occur. The local access is a shared resource between our end users and the users of the whole sellers.
Lamarre: To get back to the topic of the various products offered, you have 6 different service levels. Can you provide us with the information? Also, based on your experience, and the surveys you have done, do you think that having economic measures, and giving customers the choice of what to use, can you create a certain fairness, or are there improvements still to be made?
VT: Our satisfaction results from the survey show that things are working. Our main concern is to show that we will still have the flexibility needed to adapt.
Lamarre: In regards to privacy, there is 1 method which raises concerns, DPI. You have said that you want the flexibility to use the history that is collected, even if it is not used right now. Does this mean that you do not currently keep information?
VT: When looking at a packet, to redirect it, there is no impact whatsoever on privacy.
Lamarre: Some people have a contrary opinion. My question is what would you do if you decide to you this DPI technique, how can you ensure
VT: The way to ensure our customers is to never touch the content, and I have not heard anything in the description of practice by other ISPs that there are other providers that are looking into the content. There is no concern of privacy, in my mind.
Katz: In regards to TPIA, do you currently offer TPIA services?
VT: Yes, quite a few.
Katz: Can you segregate or distinguish traffic between wholesale users and retail users.
VT: At some point we segregate it, to "hand it off" to reseller.
Katz: Do you know if your resellers use a disproportionate amount of network capacity? Can you let us know if there is a proportionality concern?
VT: Because we charge excess use charges, we should have data on that. In our traffic analysis, we look at it globally. To answer your question, I would have to analyze the data more closely, but I am not aware of any significant differences.
Katz: Theoretically, you do have sufficient financing to fund growth in network, because you charge people more if they go over their cap. Similar to Rogers, you plan 6 months out. So, the only reason you would not be able to service demand is if you under-forecast demand.
VT: The total amount of money that is collected from excess uses charges is far, far below what would be needed for network investment.
Katz: Under what reasons would you engage in traffic management in the future? The only reason I can think of is if you under-forecast demand.
VT: The whole process of determining capital expenditure is not a uni-dimensional decision, with many factors being balanced. We try to spend smartly.
Katz: You will only build capacity if it is economically viable. Ultimately, you make decisions to grow network to serve customers, and provide a return to shareholders.
VT: We have an eye on the competitive marketplace. You are portraying network management as something you do when there is a failure. That is not how we see network management. Network management and congestion management are good things, in and of themselves. If we can create a new management technique that allows us to save $10s of millions of dollars, that allowed us to offer cheaper services to our consumers, that would be a good thing.
Chairman: You said that when one end user makes an impact on the network, that impact is felt by all end users. If you resell in a fixed quantity, how do the wholesale customers affect your network? Do you not have a "safety switch"?
VT: The TPIA model is not one where you sell to resellers an "amount" of capacity. It says to the reseller go find new users, and we’ll treat them like our own end users.
Chairman: You have no way of restraining wholesale users?
VT: There is not a chunk of bandwidth set aside for them. The key thing to remember is that the access network is shared. In a moment of congestion, their end users and our end users are equally, temporarily affected by that congestion.
Chairman: Is this reality by necessity or by design? Could you avoid this, so that the customers of resellers could not congest your network?
VT: It may be theoretically possible, but we are not aware of it being done anywhere, and we would be concerned with it being done. In a shared network context, by setting aside guaranteed bandwidth, we would be setting up two classes of end users, which would degrade the service of our own end users.
Shaw Communications Inc.
For Shaw, Ken Stein, Senior VP, appearing with others. Shaw stated they respect the ability of ISPs to effectively manage networks. They said they constantly invest in this highly competitive environment. Shaw noted that even with significant investments, they still experience network congestion challenges – especially as a result of P2P traffic which, if left unmanaged, would consumer all the available capacity. Therefore, Shaw notes, they must continue to combine investment with appropriate and necessary network management strategies. They noted that the Telecommunications Act already prohibits ISPs from granting an undue preference or from interfering with the meaning of a communication, and the existing regulatory framework is working effectively to protect consumers.
Shaw says that if they did not deploy traffic management strategies to address network congestion, there would be a severe degradation in quality of service. As recently as February 2008, an internet chassis was without proper traffic shaping for over a month, and the result was a major congestion problem with a significant increase in customer complaints. Shaw stated that they use DPI technology to shape upstream P2P traffic because this is the most effective measure to address the bandwidth consumption of P2P applications on their network.
Shaw noted that their traffic management devices operate 24/7, but the devices shape traffic only during periods of congestion, where the devices automatically detect P2P applications and slow the upstream traffic. Shaw said it is acceptable for individual ISPs to select their own network management strategy that may include a combination of approaches such as DPI, bandwidth limits, excess bandwidth usage charges, time of day pricing, caching, increased capacity and any other practices that do not breach either ss. 27(2) or 36 of the Telecommunications Act. Shaw noted that their DPI traffic shaping does not influence the meaning or purpose of the communication.
In conclusion, Shaw stated that they believe that we should rely on market forces to the maximum extent feasible and to interfere with market forces to the minimum extent necessary. The retail ISP market in Canada is very competitive, and no ISP operating in that market can afford to put itself at a competitive disadvantage by blocking customer access to content or rendering applications inoperable on their networks.
Chairman: How do you define congestion?
Shaw: We define congestion in a variety of ways. For P2P, we allocate a certain amount of bandwidth per user, and when that bandwidth is used up, our automatic devices kick in.
Chairman: You traffic management only comes into effect when the congestion gets to a certain point?
Shaw: Yes, although it is always on, it automatically begins to have an effect when congestion becomes an issue.
Chairman: Does you shaping correspond to the "peak hours"?
Shaw: There is a correlation between the traffic management and the "peak hours", but it varies from region to region. We do this on a congestion basis because we operate over 5 time zones.
Chairman: We have had evidence from some that actually identifying P2P traffic is very hard.
Shaw: We believe that the signatures left by P2P data are very accurate.
Chairman: Do you have evidence to show that you are capturing P2P accurately?
Shaw: Yes, we can give it to you.
Katz: Is the component that is public internet segregated?
Shaw: The networks are segregated, and we do not fluctuate them based on demand.
Katz: is there a document to look at which lists steps to take when there are congestion issues, or are they dealt with on a step by step basis?
Shaw: We do have a system that we use. We also try to make customers aware of their over usage, because often they are not aware of them. Our traffic management principles do not become more aggressive than the standard we have set.
Katz: You are very concerned with market forces, but what you are saying is "let us do what we feel is necessary", which is actually carrier driven. Can you explain why the economic vision is not what you are trying to achieve?
Shaw: That is not what we are saying. In a competitive market place, consumers do have a choice. If consumers are unhappy for whatever reason, they can move to another ISP. I think our customers are very happy with how we are operating. At the present time we have tried to fix the congestion problem. What we are able to do right now is satisfy P2P users, so that they can have maximum use of the system, when the system is available to them. We are trying to satisfy, to the extent that we can, all users using our network. The price would be so excessive for
Katz: What you are saying is some customers are subsidizing other customers (i.e. those who use P2P).
Shaw: That is not what we are saying. This is a business, and nobody is subsidizing anybody.
Katz: You said your customers have choice. Do you offer bundles of service, where users lock into long term contracts?
Shaw: We bundle products, but none of them are bound by long term contracts.
Katz: We heard last week that there are certain services that are not application specific, that hog the network. How do you deal with the questionable ones, at times of congestion; do you do what you need to do to protect your network?
Shaw: We always try to protect the customer; if we are not sure, there will not be throttling applied.
Katz: Do you know what the traffic volumes are for wholesale customers, as opposed to retail? Are they disproportionate (i.e. do wholesale end users use more than retail customers)?
Shaw: We can get the information for you, but I have not heard of the use being disproportionate.
Katz: You have said that you only feel required to disclose when changes would have a "material and adverse effect on the operation" of the competitive ISP. Can you define it for us?
Shaw: We would make the decision as to what "material" means. We view this as the same as any changes in network management. Wholesale traffic is the same as our own traffic, and we wanted it handled in the most efficient way possible.
Katz: You have said that the CRTC should not put the significant benefits at risk by regulation which could increase prices and stifle innovation. How would this be the result?
Shaw: What we mean by that is that we have to respond to the market, and that regulatory regimes that require prior approval generally are not amenable to that. We believe that the current framework is the correct framework.
Katz: You have said that "blocking content" might breach s. 36. Under what conditions might it not breach s. 36?
Shaw: The intent of the comment was that you do not know under what circumstances a carrier may block particular content. For example, something which is required by law, or a public policy objective, may not breach s. 36.
Katz: What would be involved if you had to follow the FCC edict on Comcast? Could you incorporate this into your submission to us?
Chairman: We have heard that Comcast engaged in session management? Is Shaw?
Shaw: No we are not. Shaw will slow down packets, but we do not act in the way Comcast did. Comcast forged packets, and this is a very aggressive and noticeable experience for the customer. We have not done what Comcast did, so why would we have to do what Comcast had to do? We do not have the same issue to resolve.
Chairman: What we want to know is what options you would have, and what it would take to address the concerns the FCC outlined in the Comcast decision.
Lamarre: In your presentation, you have said that ISPs should only be required to tell wholesale customers as to traffic management. How, then, are retail customers supposed to be able to make an informed decision as to which ISP they use?
Shaw: Customers make decisions based on service. If they are unhappy, they can move to another ISP.
Lamarre: However, if consumers are not informed about traffic management, then how can they make an informed decision? How can this be considered market forces?
Shaw: People do comparisons day by day amongst competitors. Also, we do not have long term contracts, so consumers can always switch. Also, dealing with speculative issues, we could do a better job in dealing with issues that people are not sure about. However, we should do more in informing customers as to what we do for their benefit. Shaw said that they could do more to inform their consumers as to their traffic management.
Lamarre: How does the slowing down of a file not interfere with the content?
Shaw: What the customer chooses to do is download a file, but to do so, they must also allow uploading.
Lamarre: Consumers are quite savvy. They actually do make the choice to upload as well.
Shaw: We have heard that these types of traffic management do raise privacy concerns. How can you reassure your customers that you are minimizing infringement?
Lamarre: In Canada we have a very strong privacy framework, guaranteed by law. Shaw has undertaken to respect those privacy rights, and are sensitive to privacy issues. We have a privacy statement on our website. We do think, however, that the current legislative measures in place are very strong protections. Customer information belongs to a customer, and we only use it to manage our business within the law.
Denton: Many are coming to the conclusion that application specific measures are needed for the management of networks. The OIC, CIPPIC say that this is wrong, that you need to achieve this result by another method. Are you sensitive to the concern that the exercise of application specific measures may give rise to reasonable concerns about the future of the internet?
Shaw: These people say that the internet invents itself. We do not know what people are going to be able to do with the network. Our interest is ensuring that the system works. We are concerned with dealing with congestion, and our total focus is to create a network that allows people to use it the way they want. The type of applications that win are always a surprise, and this is the kind of environment that we want.
Denton: So, then, P2P applications seem to generate a lot of congestion for carriers. If this is going to be rationally managed, is there a way of solving this problem without affecting the right of creators to create these types of applications.
Shaw: We in no way are restricting people’s right to create new technologies, we are doing the opposite. Our traffic management is not meant to restrict P2P, we are trying to allow all other protocols to flow in an efficient manner. We actually give a disproportionate amount of "the pie" (i.e. we allow 30% of bandwidth for P2P, even though only about 10% of users use it).
Denton: The notion that the growth of traffic on the internet is much slower than had been predicted a decade ago, and that ISPs have the ability to keep up with the traffic, has been presented to this panel. What are your views on this?
Shaw: The growth of the product differs, based on the year. Some years it has doubled, tripled, and right now it is growing at a rate of 50%. If things stayed at this rate, we may be able to deal with it. However, things change. We did not predict that video file sharing would become as dominant as it is today.
Denton: What is the balance between appropriate prices and traffic management?
Shaw: We want to balance prices to keep ourselves competitive, just as we try to balance our traffic management mechanisms. I am not aware of any complaints made by Shaw users based on traffic management?
Denton: What controls it best, prices or traffic management?
Shaw: Pricing is more important than anything we do with traffic management, because users are generally not aware of traffic management policies.
Molnar: Is targeted traffic management more expensive that 24/7?
Shaw: No it is not, not for us.
Molnar: Rogers said that putting in place a dynamic traffic management approach could result in unpredictable customer experiences. What are you views?
Shaw: We do not believe that customers see any difference in their internet use as a result of traffic management. We do not think they are aware of whether or not it is on or off. Most customer complaints do not deal with this issue.
Molnar: Did I hear you say that when there is congestion, P2P can be throttled back to 80kbps?
Shaw: This is the minimum, and will allow P2P to work relatively well.
Molnar: I want to ask about a particular suggestion, about whether ISPs, along with advertising speeds, to advertise throttled speeds. Do you have comments on this, as a means of full customer disclosure?
Shaw: Even if we did, this would not be very useful information for the customer to have. It is not what the customer is looking for in the product. Customers want to know about download speeds, customer service, etc. They are not concerned with the throttling of P2P traffic.
Molnar: If in the future you needed to throttle other things, such as downloads, would you alert customers?
Shaw: That will affect directly the quality of service the customer receives. That information would be made aware to them by their use of the service.
Molnar: There is a difference between being made aware of something before hand, and discovering it as the result of using the service.
Shaw: We do not disagree with that. We want to increase speed. We want to increase downstream capability.
Chairman: Question about the shared nature of the network with TPIA.
Chairman: Can you segregate time sensitive applications, like Skype? How can you tell what applications are time sensitive?
Shaw: We do not slow down applications that are time sensitive. We have technicians that have developed methods of telling the difference between time sensitive applications, like Skype.