said by viperlmw :

I got lambasted here by an ex-mod for suggesting that QOS has the potential to be useful

said by NetAdmin :

The "potential" of QoS/prioritization to "useful" is limited.

said by viperlmw :

I agree. I have always agreed. The primary benefit to implementing QOS IMHO has always centered around it's use in assuring time sensitive packets get thru as needed in applications such as streaming video/audio and VOIP. No amount of network upgrades or other management tactics, reasonable or not, will ENSURE availability of bandwidth to these types of applications in all segments of the network at all times. Malfunctions happen. Unanticipated capacity utilization occurs. QOS can be used to assist time sensitive applications under these and other circumstances.

said by NetAdmin :

However, you seem to overestimate the ability of QoS and underestimate its abilities to have undesirable effects. I went and found the conversation you had with the moderator many moons ago. Yes, QoS is very useful in helping out time sensitive applications, but it can also create issues, as was hashed out in your conversation with the moderator. The moderator was correct in the statement that prioritizing classes of traffic has an effect on other classes of traffic and it can be negative.

said by funchords :

All traffic is forwarded at the speed of light.

Big caps so that you know it is true: THERE IS NO SUCH THING AS PRIORITIZATION. There is only "deprioritization."

In order to allow certain packets to get ahead or cut in line, network devices slow down and fragment everything else.

Comcast isn't very interested in proper DiffServ because it doesn't allow them to run their networks under extended congestion conditions. They're trying to avoid upgrading the network as often as they otherwise would have to. Unfortunately, there is no alternative for that problem.

said by NetAdmin :

First and foremost, you are correct in the assertion that providers can never have 100% bandwidth available, 100% of the time. Those of us who work in the industry work our hardest to ensure that we can get as close to that as possible without breaking your or our bank.

However, you seem to overestimate the ability of QoS and underestimate its abilities to have undesirable effects. I went and found the conversation you had with the moderator many moons ago. Yes, QoS is very useful in helping out time sensitive applications, but it can also create issues, as was hashed out in your conversation with the moderator. The moderator was correct in the statement that prioritizing classes of traffic has an effect on other classes of traffic and it can be negative.

There are other limitations to QoS, which I think were touched on in that conversation you had with the moderator. If QoS is not maintained from your system to the target host, its usefulness is severely limited. Here's an example - All providers love peering points, but peering points are oft congested and there is no QoS at them. So a VoIP call that maintains a good flow in the first mile due to QoS can be totally decimated at a peering hand off. And then there is no guarantee that your ISP's transit provider or peer will even respect the QoS assigned to your packet anymore.

If you and others are going to advocate QoS, it needs to be advocated across the board, on up to transit providers. However, before that position is taken, people need to understand what type of impact that would have...

said by NetAdmin :

I'd like to know where you found that information...

said by funchords :

Think about it -- write it in psuedocode -- how WOULD you give traffic priority? You wouldn't -- you can't. You can only slow everything else. It's like how a firetruck or an ambulance gets priority, everyone and everything else pulls over to let it through.

said by NetAdmin :

Part of the problem is that Comcast and others have been pulled into the battle of the cablecos versus the telcos, resulting in insane speeds for ridiculously low prices.

said by funchords :

That said, Comcast has been offering 6 Mbps since mid 2005. Moore's Law and Cisco's worldwide bandwidth trends both say that our speed should have doubled about every two years. It's now mid 2008 and we're looking at 6 months to a year before Comcast's next speed upgrade.

said by funchords :

For example, I cite the mighty uncitable Wikipedia: http://en.wikipedia.org/wiki/Moore%27s_law#Other_formulations_and_similar_laws

said by NetAdmin :

Sadly, short sighted investor mentality and the idea that you only need just enough to keep up with the competition is a mainstay of the American way of doing business.