Keeping up with customer bandwidth demands is a big challenge for service providers today. Throwing bandwidth at the problem is not the solution. Instead, bandwidth management policies that moderate peak-time usage and tie revenue to bandwidth used can help operators stay ahead of the curve while getting the best return on their bandwidth investment. 

To gain more insight into the bandwidth dilemma, we talked to Bob Hunt, a Broadband telecommunications technologies and management consultant. Bob has over 24 years in the North American and European Cable spaces helping clients architect, build and manage the technology relationships in complex Video and Data network environments.

Bob was recently Vice President of IP Engineering at Charter Communications where he led the development and management of DOCSIS, IP Backbone and regional cores, VoIP, Video and Data Labs as well as Capacity Management and planning. He has also worked with numerous technology vendors to assess, design, promote and plan various Data and Video technology products and their delivery to market.

Obviously broadband in all its forms has changed dramatically over the years. What is the main issue providers are now faced with?

Hunt: The main issue, the biggest issue is bandwidth consumption. When you look back at the beginnings of broadband, the bandwidth consumption back then was nothing compared to what it is today. In fact, no one really gave it much thought. Now, you have real-time entertainment applications, such as Netflix, Hulu, Amazon and peer to peer sharing, that are consuming a great deal of bandwidth. In the process this is creating a tremendous cost to the operator as they try and keep up with bandwidth demand. This bandwidth squeeze is forcing operators to purchase more and more bandwidth and those costs are escalating, even if you can get it, which isn’t always easy to do. And if you look at the cost of that bandwidth, breaking it down based on the average revenue per user, those rising costs obviously are affecting the profit margins for operators.

That’s the one big issue I’m hearing every day. The other important issue impacting operators is the effect of the dwindling bandwidth on the subscriber’s experience.  Because of the type of networks we run, a “top talker” who is downloading and uploading a lot of data may very well be impacting another user’s experience. We always want to be cognizant of that, and that’s one of the reasons why there are broadband caps out there. We want to ensure that we can put a cap in place to mitigate demand if need be. It also allows the ability to upsell subscribers into additional data tiers.

In the past providers have used deep packet inspection (DPI) products to manage bandwidth. How has that changed? 

Hunt: DPI was useful for addressing bandwidth usage when peer-to-peer traffic was the only real bandwidth problem providers were dealing with, because it can manage traffic based on what application a subscriber is using. For example, if several users were sharing files on BitTorrent or Kazaa, you could manage only that traffic to solve the problem. Today, it’s the average user that needs to be managed because that’s where the bulk of the bandwidth use is coming from. Plus, managing bandwidth based on the application doesn't fit within the FCC’s network management rules because it isn't net neutral.

While DPI is still useful for getting an idea of what specific traffic is flowing across a network, it's a costly solution for managing the traffic on that network. First, you need an expensive box to sit in-line in the network for each access point. Plus, you have to upgrade those boxes as bandwidth increases. A better solution is a system that uses IPDR to centrally collect the usage from each CMTS and then direct the CMTS to restrict usage based on things like the subscriber's package, the time of day, or congestion on the network. By leveraging IPDR there’s no need for additional equipment in each headend, and no upgrade costs as overall bandwidth increases.

What exactly is IPDR?

Hunt: That stands for Internet Packet Detail Record. It's part of the DOCSIS standard, and used to collect usage data from the CMTS. There are bandwidth management products that will collect and process that IPDR usage data from the CMTS, and they can tell the CMTS to enforce the provider's predefined bandwidth use policies.

What about monthly bandwidth quotas based on total usage. Does that help solve the bandwidth problem?

Hunt:  Not really, if that’s the only bandwidth monitoring being done. What you wind up with is a lot of usage early in the month, and it really doesn’t do anything to help moderate usage during peak periods. So, it doesn’t do much to lower your costs or improve the customer experience. One of the challenges that operators have is that they must purchase enough bandwidth to handle peak demand, so managing peak demand is critical for maintaining a good customer experience as well as keeping costs in line. You can have monthly quotas, but you need to combine that with daily bandwidth management.

High bandwidth use has become more of a day-to-day function, or perhaps that should be said as an evening to evening function. For example, you start to see usage jump up at 6 AM, and it gradually ramps up until 5 PM. After 5 the peak begins to accelerate substantially to 8 or 9 PM when you’re really seeing the peak of the bell curve until midnight. Then, at midnight it starts to drop off.

Also, having a system that can look at network congestion in real-time and apply bandwidth restrictions whenever needed is important, because there may be other factors that affect bandwidth regardless of what time it is.

 Continue reading....

 So, effective bandwidth usage management needs to take into account big events on the Internet that a lot of people may watch?

Hunt: Yes. Now the Super Bowl-type events and things like that are usually multi-cast sessions, so a lot of people are joining on to the same multi-cast which ultimately is not consuming that much bandwidth within the network. What does consume bandwidth, for example, is say Apple coming out with the latest update for the iPhone and everybody jumps on one minute apart. It’s a couple of gigabytes to download, and each user is starting up an individual unicast to be able to download the updates. Something like that can have a huge impact on the network.

Some providers will throttle users regardless of the time of day or month if they are using a lot of bandwidth. Is there a better way?

Hunt: Yes. Let’s take, for example, the typical family, mom and dad, two children, having multiple PCs or devices and one person is watching a movie and another is downloading files, yet another is listening to music. Individually, they are not consuming a huge amount of data, but collectively they are. During the day, it’s not really a problem because the pipe’s large, so you could let them use whatever they want. Let’s say they’re on the lowest tier package, their experience is satisfactory for them and they’re not impacting the other users in their area.

Now imagine that same family around 5 or 6 PM when everyone else in the area starts to come online. If you’re monitoring, you begin to see the growth and consumption of bandwidth within their service group. Looking at our sample family’s amount of bandwidth consumption, you realize that this particular household, compared to everyone else, is not within the normal range of consumption. As other subscribers begin to come online, the high consumption users, like our family, will start impacting their neighbors.

Now would be the opportune time, based on their collective usage during this peak time, to throttle them about 10%. Throttling the users isn’t going to have an impact on their ability to do what they’re doing. It may slow down some tasks a little, but the user experience is going to be close to what it was during the day. The difference is how this small amount of throttling will benefit everyone else in that service group. Overall, that small percentage, especially if you throttle 10% across the service group, can ultimately benefit everyone during heavy-use peak times.

As you said, you want to do any bandwidth management in a net neutral way. Would you do this across the board whether it was Bit Torrent or Netflix or something similar?

Hunt: Yes, you would want to do it across the board. FCC rules allow for bandwidth management if it is neutral, and not punitive against a particular application or Web site. So, whether someone is surfing the Web, downloading a file or watching a movie, they would see the same reduction in bandwidth. The good thing about most of the service slowdowns based around Netflix and similar services is that they are able to handle throttling. What will ultimately happen is that the application will realize it’s not getting as much bandwidth as before and basically, with Netflix for example, the picture quality will soften. In fact, even if there is plenty of bandwidth locally, if Netflix is overloaded, the same thing will happen.

And, there is another aspect to this peak/off-peak usage. It’s not as simple as designating evening hours as throttling hours and daytime as no holds barred. When an operator wants to optimize bandwidth usage and revenue; he wants to fill the down time or “empty” bandwidth time during the day and the best way to do this is with commercial customers. Commercial customers are primarily day users but collectively they can tie up a lot of bandwidth. Therefore day monitoring is also necessary to ensure those commercial customers are kept within their own data usage tier and not impacting residential customers and vice versa.

And commercial customers are going to be paying a higher rate for reliability and consistency, and you need to be able to provide that?

Hunt: Absolutely. You’re going to be selling them larger bandwidth packages, as well, and the satisfaction rate of these customers can really impact an operator’s revenue. So, it’s important that any bandwidth management system be able to handle differences between service groups, such as residential and commercial, and apply different policies based on packages that are purchased.

Let’s talk about the usage data and how it’s collected. For a cable provider, is IPDR the easiest way to collect usage since it’s a DOCSIS standard, and what about accuracy?

Hunt: That’s correct. IPDR pulls the data right from the CMTS, so it’s highly accurate in that regard. We used to use SNMP to track usage, but it is not as accurate because of how it is collected. For effective bandwidth management you have to collect usage real time with IPDR, and you can actually pull the data as often as you want (every hour, ten minutes, etc.). It depends on how close to real-time you want to get for service management. IPDR can collect a lot more information than SNMP.

Any thoughts on providing usage information to end users and the need for that to be accurate?

Hunt: If you’re going to apply bandwidth policies or charge for usage, I think it’s important for you to provide a way for customers to view their usage, and for that usage to be accurate. I think as you look at monitoring tools, Comcast is kind of the classic model right now. You can see how close to your limit you are, and you can go in there and check it whenever you want. It is a valuable tool.

Any thoughts on how providers should be charging for usage to recover some revenue?

Hunt: Obviously, it’s one of those situations where if you want to drive a luxury car, you pay for a luxury car. You can look at it as lanes on the freeway with the luxury cars getting the faster, wider lane to drive in. But you want to make sure everyone stays within their respective lanes.

Bandwidth has a cost to it. If you’re willing to pay for as much bandwidth as you need, (the luxury car), then the provider puts you in the fast lane. If a user doesn’t want to use a lot of bandwidth, but if you still want to get on the freeway at a reasonable speed, the provider slots that user into a different, lower cost lane.

With networks today, you can’t sell everybody a Rolls Royce. The network “fast lane” wouldn’t be large enough to keep up, using the freeway analogy. Conversely, you can’t have everyone in a VW Bug all in the same lane either. You want to sell everybody the speed that they want, while keeping up the consistency in the user experience with all of those speeds, especially at the peak time. The operator needs to ensure that at each level every customer is getting full quality for their money.

A lot of providers will price their tiers on speed, but even at 3 Megs being the lowest priced tier, that’s plenty fast enough to stream movies, etc. Couldn’t it be true that these users are using the most bandwidth?

Hunt: Absolutely. I work with an operator in Connecticut, and there is a guy, who’s in the 5 MB tier, and last month he downloaded 3.2 TBs, and he was in the lowest service tier.

So, because of that, it would be important to have usage quotas for people in the lower service tiers because everyone is going to gravitate toward that lowest price.

Hunt: Right.

So, if a user wants to do HD streaming and cut the pay TV cord, he could do that with 3 Megs, and they’re going to kill your network.

Hunt: Yes and that’s when the trouble starts, because everyone else is impacted by that subscriber. That’s when you have to implement caps so you can throttle those users who are joyriding in the “luxury car” that they aren’t paying for.

Do you have an opinion on the pay-as-you- go service offerings? More specifically, where an operator could offer a higher level of bandwidth when it’s needed or some type of pre-pay system that when you’ve exhausted all of your bandwidth, there isn’t any more left?

Hunt: One thing I think you have to look at now is the level of complexity of the billing systems and being able to deal with that model. But yes, if someone could just put in a credit card and buy a certain capacity for a certain amount of time, I believe as a service provider, we’d have to be flexible to give them what they want when they want. Operators are going to have to start thinking about broadband differently since there are such big usage amounts.

A classic example would be a subscriber who is on the low tier all year and then the holidays come along and they buy the kids iPads and gaming systems, and they want to download everything all at once. Then, after that it calms down and kids go back to school, so they’re perfectly fine going back on the lowest tier because it’s all they need. So the pay as you go scenario would solve a lot of problems, not to mention earns extra revenue for the operator.

The best approach for bandwidth management is for the operator to ensure the latency is low, and user experience is extremely high, and to make the needed investments in the network to ensure that. The whole reason behind usage-based service management is to ensure that the operator is investing in terms of network assets appropriately to continue to grow the network, and pricing the service in a way so they can recover that cost from those who are using the most resources on the network.

Don’t forget to download the white paper on Best Practices for Implementing Broadband Usage Management.