QoS and Switches

[Serpentarius]

Hi, i've been pounding on the internet on the Quality of Service (QoS) articles and i couldnt feed the logic bout QoS. How it works? Does it require some specific software/driver to function?

I' interested to acquire more knowledge on the networking. Care to discuss?

[yankee]

Pretty interesting stuff.

http://en.wikipedia.org/wiki/QoS

http://en.wikipedia.org/wiki/IEEE_802.1p

[rogard]

QoS - can be deployed in various different forms that probably why your having such a hard time nailing down a definition

I would suggest grabbing this:
http://bandwidthcontroller.com/trafficShaperXp.html

It is true that traffic shapping and QoS are not the same thing, however traffic shaper I think is a nice practical way of getting the uses and deployment in to your head.

The idea of queues and pdu tagging (cell/frame/packet etc) is pretty much present which ever platform/layer shapping is deployed.
and the above app should help with that.


p.s. i know its temping to play with the "QoS packet scheduler" install by default on xp, however it is leading you down the wrong path as it is of little use in a domestic qos deployment.

[ARC1450]

Check on the cisco site. It'll have a bunch of information on it and how it's deployed.

[Serpentarius]

Check on the cisco site. It'll have a bunch of information on it and how it's deployed.

cisco seems too heavy for me feed ...
ok i'll read it later

[ARC1450]

Well (and pardon me in my inebriated state), the big thing about QoS is that is controls bandwidth. There are several mechanisms to control bandwidth, and those are drop priority, shaping, and simply priority (IMHO). Keep in mind when I say this, I'm plastered off my ass.

With drop priority, generally what you can do is remark traffic so that the less important it is, the more likely it is that it will get dropped. Shaping is generally employed with drop priority, and it includes allocating a certain amount of bandwidth, and once that bandwidth restriction is reached, packets either get marked to a higher drop priority or dropped completely, depending on how important they are. There's also priority queuing, and this is simply setting aside a certain amount of bandwidth for a single application. This is really only good and useful on low-bandwidth (<2M) links.

Keep in mind that there are, for lack of better terms, derivations of all of these (IIRC, which I probably don't). With some switches, you can specify queue length and queue priority, thus dropping anything out of lower priority queues by what is called tail dropping, and that is the last packet in a queue gets dropped if there is no more space in the queue essentially.

And yes, Cisco's QoS explanation is heavier than lead, and at times extremely convuluted.

[edit]I should mention though, that these are generally something resembling border or core QoS techniques. For switches (such as access layer switches, or those closest to the user), generally you'll either trust the QoS you're receiving (bad idea many times), or you'll remark the traffic to the proper priority depending on where it comes from. Most used are DSCP values, at least in sco equipment.