WS-MINI QoS

[joeyr-stc]

I believe I read that the QoS Queues are from high to low (7 - 0) with high (7) being higher priority and low (0) being lower priority.

With that in mind I have a few questions...
1) What method is used for the queues? (FQD, etc)

2) Is the separation of two queues by -1 the same as a separation by -5 if there are only two queues?
In other words... If I have only two queues defined, would queue 1 set for level 7 and queue 2 set for level 6 be the same as queue 1 set for level 7 and queue 2 set for level 1? Would they act the same or does the distance apart effect queuing behavior?

[Stephen]

Frames are actually processed in the order of the rule sets on the QoS tab. The rules themselves determine which queue to send the matching frame too.
The rules are also processed from top to bottom. So if you have 2 rules, say 1 and 2 and rule 1 directs a frame to queue 4 and rule 2 directs a frame to queue 6. Then say a frame comes in that matches both rules, it will be directed to queue 4 because that is the one that is checked first. So in this sense, the priority of the queues is arbitrary based on your configuration.

In terms of the rules themselves, you have several options to process the frames, you can use EtherType to select which protocol frames should be, Source/Destination MAC addresses to check the origin of the frame, etc. You can also specify which ports these rules apply too as well.

As far as the queue's themselves, of course you can pick the RX and TX bandwidth limit. Then you can pick how loosely it holds to this limit. If no matter what frames directed to this queue should not be allowed to exceed the limits, set the Scheduler to 'Strict', if however, during moments of downtime it would be OK for the limits to be exceeded change the scheduler to 'WRR' or Weighted Round Robin.

[joeyr-stc]

Stephen,

Thanks for getting back with me.
What you said makes sense but still doesn't quite answer what I'm looking for.
It may help to explain how I am using the switch because I am not using it in the traditional sense for a WSIP installation.

I work for a telecommunication company that deals in VoIP. One of our biggest challenges is QoS.
When we arrive at a customer, we often have to split their internet circuit. The most common way this is done is with a small switch. Then, each party (Voice & Data) have their own WAN connection to use. The problem is, we have no control over usage or QoS. The data router could be using the entire circuits capacity and that would degrade the voice side.
That was when I discovered the Netonix switch that allows us to limit a port and therefore limit a connection's speed. This is great and works wonderfully. It is the main reason we started using your switches. The problem is, this is not really QoS, it is simply rate limiting. That is where the queues come in.
I want to assign the connection (port) to the data router a lower queue and the connection (port) to the voice router a higher queue. So, if both are competing for data, the voice port will get a higher priority treatment.

I am currently creating rules based on the default vlan because that is the only way I could grab all the packets from a particular port. I tried EtherType: 0x0800 but that did not seem to work. (See post: viewtopic.php?f=6&t=3921)

Currently...
Packets on port 6 (the connection to the ISP) on vlan 1, are assigned to queue 6.
Packets on port 5 (the voice port) on vlan 1, are assigned to queue 5.
Packets on port 4 (the data port) on vlan 1, are assigned to queue 4.
[img]g:/Netonix_QoS.jpg[/img]

Does this give the traffic on port 5 (the voice traffic) priority over the traffic on port 4 (the data port)?
Also, if I put port 4 (the data port) in a lower queue, say 1 (with nothing in queues 2-4) does that do the same thing as it being in queue 4?

I know this is a long explanation but if this works like I think it should, it could transform the way we do things for our entire customer base.

Thanks.

[Stephen]

Hey joeyr-stc,

Sorry for the wait but I wanted to be sure when I answered.

On our switches, QoS is prioritized starting from queue 0 to 7.

Also, as it turns out if you assign priority packets to queue 0 and less prioritized packets to queue 1 it will behave the same as if they were set to queue 0 and queue 7.

[joeyr-stc]

Thanks for clarifying.

However, I could have sworn I was originally told that the higher the queue number, the higher the packets are prioritized.
According to your last past it is just the opposite. The lower the number, the greater the priority.

I would definitely like to confirm that.

Thanks.

[sirhc]

Thanks for clarifying.

However, I could have sworn I was originally told that the higher the queue number, the higher the packets are prioritized.
According to your last past it is just the opposite. The lower the number, the greater the priority.

I would definitely like to confirm that.

Thanks.

I will let Eric or Stephen confirm that but you might be confusing the priority on the Power TAB which is backwards of that. I argued that myself when we did that, I thought it was backwards and lower # should be higher priority.

The QOS is built into the core so they probably are the more conventional way of thinking where the lower the number the higher the priority.

[joeyr-stc]

I would like to get confirmation from Eric or Stephen, just so I am 100% confident.

Thanks for your quick reply.

[joeyr-stc]

One other thing, It appears that the default queue is zero. (When no queues are defined, the port statistics show traffic on queue 0).
It would seem to me that the default queue (zero) would be the lowest. Using that login, 7 should be the highest queue, the queue that gets priority.

Again, confirmation from Eric or Stephen is critical to knowing 100%.

[Stephen]

I actually spoke with Eric about it before I responded and looked over the doc's from our switch-core manufacturer to be certain.

[joeyr-stc]

OK, so zero is the highest priority queue and 7 is the lowest priority queue!

Thanks.