Netonix Forums

If I enable rate limiting on a switch port, what method does the switch use to drop packets?

WRED, Tail Drop, etc?

I have a thought... depending on this answer.

mhoppes wrote:If I enable rate limiting on a switch port, what method does the switch use to drop packets?

WRED, Tail Drop, etc?

I have a thought... depending on this answer.

Method: Tail Drop

Thanks... Question:

Can the Vitesse switch core support VoQ (Virtual Output Queues). With VoQs, the ingress port will not send a frame to the egress port unless there’s room. If there isn’t room, the frame will stay in the ingress buffer until there’s room.If the ingress buffer is full, it can signal the sending port it’s connected to to PAUSE.

The other thing I'm wondering is if rate limiting 100meg APs on the switch to something like 40 or 50megabits would help with throughput issues from buffer overflow. Since the airMax 100meg gear does not support flow control, by limiting the switch port to only be able to send the maximum over the air speed (or close to it) perhaps it would help cut down on bursty issues trying to shove 100megabits down a 30megabit pipe?

QOS is planned very soon.

I met with Rory (my old CFO) yesterday to discuss how we should implement this into the firmware and yet keep the UI simple.

We are also discussing a little more configuration of the QinQ which the basics are in the BETA v1.3.0rc9 firmware.

That's good, but QoS isn't really what I was asking about. I'm talking about keeping the transmit buffer from filling up by not sending it packets if it's already full. Maybe the switch already does this?

I'm testing rate limiting at the switch port to see if it has any affect.

That being said, rate limiting the switch port makes no difference... oh well, it was worth a try.

hmmm.. Is that because the problem can happen well before the AP is saturated. I.E. The real problem is that if only one CPE is requesting data and maxing out it's connection there can still be a rush of data in to the network from upstream?

I'm trying to figure out if this is a problem on our network. We rate limit at the head end (powercode bmu) but after that point it's layer 2 VPLS from their to each tower and on to the customer even though the network is routed.

We do 20Mb packages in one part of the network but it's being fed with Exalt 11Ghz 300Mb link maybe that's why? Every thing else is 8Mb or less.

I am just getting ready to head into RF Armor.

I do not ever have any issues Matt but then again I run Flow Control from the AP to the "Tower" Routers plus I have a Static LAG from the Switch to the Router to spread out the Flow Control Pause Frames from the AP's on each tower across 2 interfaces and I am pushing Pause Frames to a Routed Way Point at each tower.

This is the path to the internet from my house:

Level 3 Fiber Internet=>Router=>BWMGR=>Router=>AF24 Link=>WS Switch & Router=> AF24 Link=>WS Switch & Router=> AF24 Link=>WS Switch & Router=> AF24 Link=>WS Switch & Router=> AF24 Link=>WS Switch & Router=> AF24 Link=>WS Switch & Router=> Legacy M5 AP 30 MHz channel DFS=>NanoBridge M5=>My House

I share the AP with about 10-15 other customers. AP is about 2 miles from my house.

CLICK IMGAE BELOW TO VIEW FULL SIZE

Not to derail...what do you use for BWMGR, Chris?

Well my house is not being limited so my rule in the BWMGR is just collecting stats for me.

But http://www.etinc.com

I have run for weeks with it removed when it broke and my internet was the same so it is not a magic bullet

The secret formula is:
A router at every tower.
Flow Control from AP to Router.
Router MUST support Flow Control.
Use a LAG between your switch and Router to spread out the pause frames.
And a pinch of salt.