Showing posts with label deep packet inspection. Show all posts
Showing posts with label deep packet inspection. Show all posts

Sunday, November 29, 2009

Content Delivery Networks and Network Neutrality: Net Is Not Neutral

Much discussion about network neutrality seems to assume that the issue is bit or application "blocking," and from one perspective that is correct. The existing Federal Communications Commission rules about a users' right to use all lawful applications already prohibit blocking of legal applications on wired networks. The issue is whether those rules, and the other "Internet Freedoms" principles also should be extended to the wireless domain.

In another sense, popular perceptions are misguided or worse. There is a separate issue, that of whether it ever is permissible, for any legal reason, to shape traffic, either to maintain network performance, provide an enhanced service to a user, or create a new level of service.

Some will maintain there are other ways of maintaining end user experience aside from traffic shaping. That is arguably correct, but might cost so much that the entire consumer access pricing regime has to change in ways people will find objectionable.

Some argue that any traffic shaping of legal bits should be banned, because such practices have undesirable business impact. "No bits should have any priority," that line of reasoning suggests.

One might simply note that about 60 percent of video bits--almost universally served up by media companies--already enjoys such "unequal treatment." Indeed, that is the purpose of a content delivery network: to expedite the delivery of some bits, compared to others, so that a better end user experience is possible.

In fact, about $1.4 billion was spent in 2008 precisely to deliver such expedited bits. The U.S. market currently generates an estimated 55.8 percent of the global CDN traffic, though international traffic is now increasing at a faster rate than its domestic counterpart, according to Research and Markets.

And though video delivery historically has been the CDN staple, new growth areas include whole site delivery, dynamic content, "live" video, high-definition video, mobile and smartphone applications, other non-PC devices and adaptive bit rate streaming, Research and Markets notes.

Of the 22.5 billion professional video views served during 2009, Akamai delivered 31.9 percent, Limelight Networks 12 percent and Level 3 11.2 percent, says Research and Markets.. Additional CDNs active in the market include CD Networks, Velocix, Liquid Compass, Abacast, Mirror Image, Edgecast Networks, Highwinds, BitGravity, Cotendo and Internap, the firm notes.

The point is that preferential delivery of bits already is an established part of the way the Internet works. Private network users, especially businesses, also commonly set up traffic priority systems for their internal communications and content, as well.

The ability of a consumer end user to choose to use such services and applications is one of the implications of the network neutrality debate that often is lost. To reiterate, preferential treatment of bits already is happening on a wide scale, and for very good reasons: to preserve end user experience. Perhaps we ought not to be in such a rush to foreclose practices and capabilities of obvious value.

Wednesday, January 30, 2008

Broadband Buckets: Way to Avoid Packet Discrimination


Many years ago, wireless and wireline minutes of use were sold on a metered basis. These days voice and texting are sold by the bucket. There's now more experimentation with that sort of model for broadband access as well.

It might seem odd, but changing the way broadband access is priced at retail, using a model similar to wireless minutes of use and texting, might be beneficial for end users, not simply for Internet Service Providers.

The reason is that if a user wants to buy a bigger bucket to move more packets for peer-to-peer video, the user is happier and so is the provider, who is able to match revenue with use of network resources.

That's arguably a better solution that having ISPs deploy sniffing and packet inspection capabilities so they can inspect all packets (as happy as some solution providers would be to sell all that capability).

Since deep packet inspection has to impose some overhead and latency, the user's applications arguably should work better as well (also avoiding privacy concerns). If any user is found to be shipping around video bits in violation of copyright, there are other remedies.

That's the way enterprises and businesses buy bandwidth, by the way. They pay more money but have unrestricted right to use the bandwidth they've purchased. Pricing consumer access in the same way wireless text and voice now can be bought would allow users to make their own choices about what applications they want to use, and how much.

It isn't metered usage in any way more objectionable than buckets of minutes or texting are. And it might allow ISPs to avoid the DPI effort.

Friday, December 14, 2007

Fair Use: Tragedy of the Commons


I might not be the most popular user in defending "fair use" policies, but I have to tell you there is such a thing as the "tragedy of the commons."

Without being overly literal about it, the "tragedy of the commons" is a way of describing how free access and unrestricted demand for a finite resource ultimately dooms the resource through over-exploitation.

This occurs because the benefits of exploitation accrue to individuals or groups, each of whom is motivated to maximize use of the resource, while the costs of the exploitation are distributed among all those to whom the resource is available.

As a westerner, I'll illustrate the problem by pointing to the history of conflict over grazing and water rights. Assume you are a cattle or sheep rancher, grazing those animals on open range that actually is owned by the U.S. government. Assume the market for livestock is good. Each rancher then has an incentive to add animals to the herd, increasing the intensity of grazing. At some point, there isn't enough grass to support all the animals.

Now Internet access is a shared resource, by definition. If you use a cable modem, the actual bandwidth is shared by a large number of end users. If you use Digital Subscriber Line, the sharing happens further up in the network, but the resource still is shared. "Oversubscribed," we like to say. One never provisions enough bandwidth to meet the full theoretical demand any single subscriber might use.

Basically, designers use statistics to provide enough bandwidth to meet average demand, at average times of day, and day of week, to meet the demand created by users who actually are online and using the resource at any given point.

But those statistics are based on "typical" demand. So what might be typical? For a consumer user, somewhere between one and three gigabytes of use in a month. My business use--and I am on the Web all day from roughly 6 a.m. to 8 or 9 p.m.--runs about 2.5 Gbytes a month, typically.

There always are a small number of users who "graze their cattle" vastly more extensively than the rest, creating something that might be less than a major "tragedy of the commons" problem, but clearly consuming enough bandwidth that user experience for all the other users paying the same amount of money is degraded.

"T'aint fair." There's a solution for very-high usage: buy a business plan that really offers "all you can eat" bandwidth at the level you require.

At Qwest Broadband, for example, the illustrative volume that really is excessive for a consumer user might be:
• 300,000-500,000 photo downloads in one month
• 40,000 to 80,000 typically sized MP3 music downloads in one month
• 15+ million unique e-mails each month
• Online TV video streaming of 1,000-3,000 30-minute shows each month
• 2-5 million Web page visits (approximately one every second, 24 hours per day)

Those of us who have jobs, spend time outdoors, play sports, garden, ski, raise children, go shopping, read books and so forth really don't have time to consume that much data in a month.

Some people might have to do those sorts of thing for work, but that's the point: buy business bandwidth that clearly is sold with the understanding that if you want to push the network that way, you pay more for the privilege.

So long as access bandwidth is a shared resource, there will be a "freeloading" or "tragedy of the commons" danger. Good citizenship, good manners and good neighborliness requires a little respect for other people here.

I fail to share the "outrage" of people who think they should be allowed to overgraze the commons. Nobody has a "right" to impose those costs on the rest of users who "play nice."

"Tokens" are the New "FLOPS," "MIPS" or "Gbps"

Modern computing has some virtually-universal reference metrics. For Gemini 1.5 and other large language models, tokens are a basic measure...