Traffic shaping

[rn4it]

I was told by our telco the issue we are having with transfer speeds going from HQ to the satelite site. Was an issue that could be corrected by configuring traffic shaping on the HQ wan router. I was told to place the traffic-shape rate 256000 cmd, under the ser 0/1 interface. This made no change, it still takes 2time as long to push data to them, then it does to pull data from them. We have a T1 going into the Telco and the Satellite site has a 256K. I have also attempted to place the traffic-shape rate cmd in ther serial 0/1.2 interface which is the subinterface that connects to that branch. Any one have any idea's?

 

[routerman]

Do you have a LAN analyser available, such as a sniffer?

If check for retransmissions, you may be dropping packets across the link, this results in retransmissions.

Are you using Frame relay, if so look at the output from SHOW FRAME-RELAY PVC, this shows how many congestion flags have been received. Best way, clear counters first then do a transfer anf check the output from this command.

Post it up if your not sure. Also look at SHOW INT S0/1 during transfer, what rates do you see?

Compare this to the contracted rate with your supplier.

 

[rn4it]

No congestion flags, No dropped packets. If I transfer a 20MB file I get approx 70Kb/s on a 256Kb/s circuit. I'll use a sniffer on Monday, I have specific times that I can test in. I working with the Supplier as well, they are stating that what is happening is on the side with the 256K it's buffer is filling up, and is sending retransmits back to the egress port which is a T1. They want us to use traffic shaping on the router to throttle back on the T1 connection.

thanks

 

[tbilan]

Traffic shaping tries to avoid congestion but it won't make your line any faster.

http://www.cisco.com/en/US/tech/tk652/tk698/technologies_tech_note09186a00800d6788.shtml

Also, theoretical maximum on a 256k link is 32kB/s so somehow you've managed to double the maxiumum possible transfer rate. Maybe your CIR is 256k and you have a 768k port?

If your application is TCP based it will slow down for you automatically. If you have a streaming UDP application you can try enabling RED but if the stream needs 384k and you only have 256k then it just isn't going to work.


Tom Bilan
TJBA, Inc.
CCNP, CCDP, MCSE & CNE

 

[rn4it]

tbilan, Nope we have a T1 into the telco clould and a 256K leaving the cloud into the satellite office. We have 0 CIR on both sides. As far as the theoretical max being 32K for a 256K link, I have heard that before. Not disputing you, it's just a little surprising. I haven't seen any literature that states that. If you have some and could forward it to me that would be most appreciated.

 

[tbilan]

256kilobit per second / 8 bits per byte = 32kilobyte per second.

Just to verify, when you say 256k leaving the cloud into the satellite office you do mean you have a DSU at the satellite is programmed for only 4 channels right?

Here's a sample of one of my frame links:

[tt]
DLCI = 370, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial1/0.370

input pkts 15445752 output pkts 12204497 in bytes 1587918001
out bytes 3241057455 dropped pkts 0 in pkts dropped 0
out pkts dropped 0 out bytes dropped 0
in FECN pkts 0 in BECN pkts 164 out FECN pkts 0
out BECN pkts 0 in DE pkts 0 out DE pkts 0
out bcast pkts 99600 out bcast bytes 6374400
5 minute input rate 1000 bits/sec, 2 packets/sec
5 minute output rate 0 bits/sec, 0 packets/sec
pvc create time 5d08h, last time pvc status changed 5d08h
[/tt]

I have a few BECNs but besides that no packets have been dropped. Is this similar to what you're seeing?

What kind of equipment do you have?

Can you post the configs for both sides?

Also do a [tt]show int ethernet0[/tt] and [tt]show int serial0[/tt] and post those too.

The more information the better we can help you.



Tom Bilan
TJBA, Inc.
CCNP, CCDP, MCSE & CNE

 

[routerman]

Your 256K remote link, is that a 256k access speed, or a 256k CIR?
So currently without any traffic shaping your pumping packets into the frame T1 at T1 speed, these are trying to leave at 256K and so your traffic backs up. Once the internal buffers in the frame nodes are full the rest gets dropped. This is why you see the re-transmits.

In a case like this I would use the following traffic shaping:

traffic-shape rate 256000 32000 0
traffic-shape rate <cir value>
traffic-shape fecn-adapt

The first line values are access-speed at slowest end, the second value is the first divided by 8

The second line uses the cir value in bit per second.

Just one point about links with CIR=0, does your frame supplier issue congestion notification in such a case?

Also is there any compression running here, either on the routers or the ens systems, this may confuse the throughput figures.

 

[rn4it]

thanks routerman, the circuit is a 256K with a 0 CIR. The Supplier states that their network is so robust that we don't need it. We have the users at the satellite site use Citrix to access applications at HQ. This way they only receive updates from the screen shots. thanks again

 

[routerman]

Oh, I work for one of them providers as well. Spend millions on infrastructure, then find we have massive over capacity that no one wants anymore.

In that case the GTS config example should suffice. It will force the T1 router to restrict output to 256kbps max.

 

[rn4it]

Thanks routerman