Network Neutrality and QoS

by on July 31, 2006 · 10 comments

There are some interesting comments that you might have missed in response to James Gattuso’s post last week about VoIP quality and network neutrality. Mike Masnick takes him to task for reading more into the Brix report than is merited.

Brix CTO Kaynam Hedayat notes that his company doesn’t take a position on neutrality regualtions, but he added these clarifications:

- Based on comments from the testyourvoip user community more than half the tests were run for pre-qualification purposes (prior to signing up for VoIP). In those cases the users did not know if they had problems or not prior to running the tests.
- Close to one million tests were conducted for this study.
- The types of impairments and degradation factors that we analyzed point to network congestion. We are further analyzing the data to understand the location of congestion (core, last mile, etc.).
- Via the testyourvoip portal we measured and continue to measure “end-to-end” VoIP quality on the internet.
- The tests are conducted between the user’s desktop to one of seven locations across the globe as selected by the user. The seven locations are connected to the internet via high BW connections without any impairments (they are monitored).

Cog linked to this Ed Felten post questioning whether QoS was a good argument for neutrality regulation. Felten concludes that it’s not–that in many cases simply throwing more bandwidth at the problem is a better solution. For what it’s worth, I find Cog and Felten’s position pretty persuasive. Guaranteeing QoS is a difficult engineering problem that’s likely to require rolling out a lot of new and expensive networking hardware across the network, if it can be made to work at all. Simply building out more capacity is likely to be a more cost-effective option, and it has the salutary side effect of increasing peak bandwidth in addition.

On the other hand, Felten, Cog, and I could be wrong. It may be that QoS can be deployed in a cost-effective manner, and that non-QoS network management techniques simply won’t give us the quality of service we need for high-bandwidth, interactive applications. Which is why we should leave network owners with some freedom to experiment. No one has a monopoly of wisdom on network design, and if anyone did, it certainly wouldn’t be Congress or the FCC!

  • http://lippard.blogspot.com/ Jim Lippard

    Global Crossing has been using QoS effectively for years. We use it in two ways–first, two allow customers to prioritize their own traffic; second, to prioritize our backbone traffic, which carries voice, video, IP-VPN, and public Internet as separate logical overlays over the same physical core. The QoS in the second case doesn’t come into play often, because we try to make sure there is always sufficient bandwidth. But fiber cuts and denial of service attacks occur, and when they do, we would rather drop packets from public Internet traffic than voice calls.

  • http://lippard.blogspot.com/ Jim Lippard

    Global Crossing has been using QoS effectively for years. We use it in two ways–first, two allow customers to prioritize their own traffic; second, to prioritize our backbone traffic, which carries voice, video, IP-VPN, and public Internet as separate logical overlays over the same physical core. The QoS in the second case doesn’t come into play often, because we try to make sure there is always sufficient bandwidth. But fiber cuts and denial of service attacks occur, and when they do, we would rather drop packets from public Internet traffic than voice calls.

  • http://www.techliberation.com/ Tim

    Does that technology also allow QoS for particular protocols/TCP streams within the public Internet traffic? And is it something that could be feasibly scaled to the Internet as a whole?

  • http://www.techliberation.com/ Tim

    Does that technology also allow QoS for particular protocols/TCP streams within the public Internet traffic? And is it something that could be feasibly scaled to the Internet as a whole?

  • http://lippard.blogspot.com/ Jim Lippard

    We don’t do QoS within public Internet traffic–all public Internet traffic is classified as best effort, and there are no standards for inter-provider QoS (though we do map customer classes of service into our own, and maintain their markings across our network and across any partner networks that we use to reach locations where we don’t have our own facilities, such as China and India).

  • http://lippard.blogspot.com/ Jim Lippard

    We don’t do QoS within public Internet traffic–all public Internet traffic is classified as best effort, and there are no standards for inter-provider QoS (though we do map customer classes of service into our own, and maintain their markings across our network and across any partner networks that we use to reach locations where we don’t have our own facilities, such as China and India).

  • http://bennett.com/blog Richard Bennett

    QoS isn’t a solution to persistent bandwidth scarcity, it’s a solution to the effects that bursty traffic (such as file downloads) have on streams with modest bandwidth requirements but tight jitter requirements. And bandwidth isn’t a solution to jitter, in many cases it makes it worse by dumping more traffic into the next segment. All you accomplish with over-provisioning is a relocation of the bottlenecks, and only that temporarily.

    The Internet is more and more a mixed-use network, and that means QoS is increasingly important.

  • http://bennett.com/blog Richard Bennett

    QoS isn’t a solution to persistent bandwidth scarcity, it’s a solution to the effects that bursty traffic (such as file downloads) have on streams with modest bandwidth requirements but tight jitter requirements. And bandwidth isn’t a solution to jitter, in many cases it makes it worse by dumping more traffic into the next segment. All you accomplish with over-provisioning is a relocation of the bottlenecks, and only that temporarily.

    The Internet is more and more a mixed-use network, and that means QoS is increasingly important.

  • http://www.abc-acupuncture.com/baxqorav tramadol

    81e31de21f46 Hi http://www.abc-acupuncture.com/baxqorav tramadol

  • http://www.abc-acupuncture.com/baxqorav tramadol

    81e31de21f46 Hi http://www.abc-acupuncture.com/baxqorav tramadol

Previous post:

Next post: