taffic shaping on Frame Relay port speed

[scjlin]

We have a Sprint as our Frame Relay provider. Currently we have six branches with port speed of 256K per each to HQ (1536K). I beleive all these connections are CIR0 from sprint. Now we are experiences some latency on our branches F.R. connection. I saw high value of DE , in FECN (branches), and in BECN (HQ) router when I run show frame-relay pvc in the routers.

Is there anyways I can do on the router improve the line speed? Can I use traffic shaping even if I have Sprint network (use port speed instead of CIR)?

Thanks!!!

scjlin

 

[cull]

ok, 2 things. "in the business", sprint frame relay is jokingly referred to as "same day frame relay service" - their latencies are HUGE compared to their competitors. 2nd, traffic shaping may indeed save your butt here. also, if you're seeing BECNs and FECNs, you have bandwidth issues, as well. Forward Explicit Congestion Notification and Backward Explicit Congestion Notification. need GTS syntax provided? i'd be glad to come with it for ya.....

 

[Jynxx]

We have Sprint Frame (iFR)...switched from WorldCom. We have NO issues with latency. It is night and day difference between the 2. Sprint builds their backbone to support 95% of port speed without congestion...thus the CIR of zero.

Regardless, what you more than likely have an issue with, is "big-pipe, little-pipe". You only have 256k at the branches, but the host port is larger than that. So it has the potential to push up to port speed 1536k to a 256k pipe. That's obviously bad.

So you DEFINATELY want to use traffic shaping on the host port for each PVC.

I would also set traffic shaping on BOTH sides to throttle down to 95% of port speed when congestion occurs.

Here is an example of what I would configure on the host port:

interface Serial0/0
bandwidth 1536000
no ip address
encapsulation frame-relay IETF
no fair-queue
service-module t1 timeslots 1-24
frame-relay traffic-shaping
frame-relay lmi-type ansi
!
interface Serial0/0.1 point-to-point
description Sprint Frame Relay to Site A
ip address 10.10.10.1 255.255.255.252
no ip proxy-arp
ip ospf authentication-key treepump
ip ospf network point-to-point
ip ospf cost 10
frame-relay interface-dlci 1
class 256k
!
interface Serial0/0.2 point-to-point
description Sprint Frame Relay to Site B
ip address 10.10.10.5 255.255.255.252
no ip proxy-arp
ip ospf authentication-key treepump
ip ospf network point-to-point
ip ospf cost 10
frame-relay interface-dlci 2
class 256k
!
map-class frame-relay 256k
frame-relay adaptive-shaping becn
frame-relay cir 256000
frame-relay be 0
frame-relay mincir 243200


A similar config can be made on the remote sides for the map class, etc.

This should help clear up your congestion. If it doesn't, you'll have to open a ticket with Sprint to troubleshoot.

 

[scjlin]

Thanks Jynxx!!

Your information is really helpful. After I exam for two hours I don’t see any FECN / BECN on either side of my routers. However, the DE still pretty high, actually the “in DE pkts” value is equal to my “input pkts” value. Also on some of routers I am still getting the “dropped pkts” (which also happen before I entered traffic shaping commands). I am thinking to use compression on my routers and see if it helps. Also, is there a way for me to determine why my lines are occupying and having so many FECN / BECN in the routers??

Thanks,

sl

 

[Jynxx]

DE is Discard Eligible. All packets are marked as DE because the CIR is 0...mine is the same way and does not cause any issues. Not sure why you would be getting dropped pkts though... They are incrementing with traffic shaping configured?

Typically dropped packets are caused by a queue overflow.

You have to remember to turn on traffic-shaping on the interface and the fair-queue turned off. Make sure your remote side looks similar to this:

interface Serial0/0
bandwidth 256000
no ip address
encapsulation frame-relay IETF
no fair-queue
service-module t1 timeslots 1-4
frame-relay traffic-shaping
frame-relay lmi-type ansi
!
interface Serial0/0.10 point-to-point
description Sprint Frame Relay to Host
ip address 10.10.10.2 255.255.255.252
no ip proxy-arp
frame-relay interface-dlci 10
class 256k
!
map-class frame-relay 256k
frame-relay adaptive-shaping becn
frame-relay adaptive-shaping interface-congestion 20
frame-relay cir 256000
frame-relay be 0
frame-relay mincir 243200

 

[scjlin]

Yes, they (dropped pkts) are incrementing with traffic shaping configured. Yes on both traffic shaping on the interface and turned off fair-queue. Also my routers do not have interface-congestion command due to IOS version.

Sh traffic-shap stat

Acl Queue Pkts Bytes Pkts Bytes Shaping
I/F Depth Dely Dely Active
Se1/2.30 89 482449 240974502 482449 240974502 yes
Se1/2.34 43 554832 191342704 554832 191342704 yes
Se1/2.36 35 397584 196759354 397584 196759354 no
Se1/2.37 98 374872 120193468 374872 120193468 no
Se1/2.80 111 284033 192274689 284033 192274689 yes
Se1/2.90 0 402869 157919607 54904 50973569 no


Thanks,


sl

 

[rcasta]

Are all ways down to your branches congested? Maybe it's only one branch hogging all up. Check on per PVC statistics and post them here. I went through the same problem as you are now.

#sh frame-relay pvc

PVC Statistics for interface Serial1/0 (Frame Relay DTE)

Active Inactive Deleted Static
Local 1 0 0 0
Switched 0 0 0 0
Unused 0 0 0 0

DLCI = 100, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial1/0.1

input pkts 209318 output pkts 5946735 in bytes 45373674
out bytes 516851820 dropped pkts 0 in pkts dropped 0
out pkts dropped 15 out bytes dropped 2181
late-dropped out pkts 15 late-dropped out bytes 2181
in FECN pkts 0 in BECN pkts 0 out FECN pkts 0
out BECN pkts 0 in DE pkts 0 out DE pkts 0
out bcast pkts 8246 out bcast bytes 2762347
5 minute input rate 2000 bits/sec, 5 packets/sec
5 minute output rate 11000 bits/sec, 15 packets/sec
pvc create time 4w1d, last time pvc status changed 2d00h

 

[scjlin]

According to our employees at the branches it seems like congested at all 6 branches. When I issues ping from my point (HQ), only branch B gave my faster ping respond 60ms. All other branches are around 200ms respond time. Following are frame-relay pvc from HQ and all branches. Also my previous replay shows traffic-shap stat from HQ.

many thanks,


sl


*****
HQ#sh frame-relay pvc

//HQ to A
PVC Statistics for interface Serial1/2 (Frame Relay DTE)

DLCI = 30, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial1/2.30

input pkts 1823340 output pkts 1897780 in bytes 220742526
out bytes 1182696699 dropped pkts 2302 in FECN pkts 0
in BECN pkts 52 out FECN pkts 0 out BECN pkts 0
in DE pkts 1823340 out DE pkts 0
out bcast pkts 57571 out bcast bytes 4644970
Shaping adapts to BECN
pvc create time 8w0d, last time pvc status changed 4d09h

//HQ to B
DLCI = 90, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial1/2.90

input pkts 1445671 output pkts 1326826 in bytes 134975439
out bytes 479752037 dropped pkts 840 in FECN pkts 0
in BECN pkts 0 out FECN pkts 0 out BECN pkts 0
in DE pkts 1445671 out DE pkts 0
out bcast pkts 57590 out bcast bytes 4646186
Shaping adapts to BECN
pvc create time 8w0d, last time pvc status changed 4d09h

//HQ to C
DLCI = 34, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial1/2.34

input pkts 2085741 output pkts 1780117 in bytes 203747778
out bytes 677918039 dropped pkts 2728 in FECN pkts 0
in BECN pkts 55 out FECN pkts 0 out BECN pkts 0
in DE pkts 2085741 out DE pkts 0
out bcast pkts 57598 out bcast bytes 4646931
Shaping adapts to BECN
pvc create time 8w0d, last time pvc status changed 1w6d

//HQ to D
DLCI = 36, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial1/2.36

input pkts 1642399 output pkts 1519938 in bytes 157843852
out bytes 691734311 dropped pkts 1930 in FECN pkts 0
in BECN pkts 4 out FECN pkts 0 out BECN pkts 0
in DE pkts 1642399 out DE pkts 0
out bcast pkts 57589 out bcast bytes 4646821
Shaping adapts to BECN
pvc create time 8w0d, last time pvc status changed 14:16:07

//HQ to E
DLCI = 37, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial1/2.37

input pkts 1639456 output pkts 1447256 in bytes 175376944
out bytes 532316814 dropped pkts 2547 in FECN pkts 0
in BECN pkts 5 out FECN pkts 0 out BECN pkts 0
in DE pkts 1639456 out DE pkts 0
out bcast pkts 57601 out bcast bytes 4647356
Shaping adapts to BECN
pvc create time 8w0d, last time pvc status changed 4d09h

//HQ to F
DLCI = 80, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial1/2.80

input pkts 879979 output pkts 931742 in bytes 119523925
out bytes 621906033 dropped pkts 164 in FECN pkts 0
in BECN pkts 0 out FECN pkts 0 out BECN pkts 0
in DE pkts 879979 out DE pkts 0
out bcast pkts 57532 out bcast bytes 4643140
Shaping adapts to BECN
pvc create time 8w0d, last time pvc status changed 1d19h

===============
//Branch A
sh frame-relay pvc

PVC Statistics for interface Serial0 (Frame Relay DTE)

Active Inactive Deleted Static
Local 1 0 0 0
Switched 0 0 0 0
Unused 0 0 0 0

DLCI = 30, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial0.1

input pkts 1908536 output pkts 1834146 in bytes 1189977353
out bytes 221655752 dropped pkts 0 in FECN pkts 12
in BECN pkts 0 out FECN pkts 0 out BECN pkts 0
in DE pkts 1908536 out DE pkts 0
out bcast pkts 61852 out bcast bytes 6239392
Shaping adapts to BECN
pvc create time 16w4d, last time pvc status changed 4d09h

================
//Branch B
sh frame-relay pvc

PVC Statistics for interface Serial0 (Frame Relay DTE)

Active Inactive Deleted Static
Local 1 0 0 0
Switched 0 0 0 0
Unused 0 0 0 0

DLCI = 30, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial0.1

input pkts 1910951 output pkts 1836521 in bytes 1191486026
out bytes 221838536 dropped pkts 0 in FECN pkts 12
in BECN pkts 0 out FECN pkts 0 out BECN pkts 0
in DE pkts 1910951 out DE pkts 0
out bcast pkts 61886 out bcast bytes 6242672
Shaping adapts to BECN
pvc create time 16w4d, last time pvc status changed 4d09h

=================
//Branch C
sh frame-relay pvc

PVC Statistics for interface Serial0 (Frame Relay DTE)

Active Inactive Deleted Static
Local 1 0 0 0
Switched 0 0 0 0
Unused 0 0 0 0

DLCI = 30, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial0.1

input pkts 1789003 output pkts 2097680 in bytes 681377371
out bytes 205635085 dropped pkts 92 in FECN pkts 0
in BECN pkts 0 out FECN pkts 0 out BECN pkts 0
in DE pkts 1789003 out DE pkts 0
out bcast pkts 61940 out bcast bytes 6107644
5 minute input rate 23000 bits/sec, 19 packets/sec
5 minute output rate 17000 bits/sec, 21 packets/sec
Shaping adapts to BECN
pvc create time 1w6d, last time pvc status changed 1w6d

=================
//Branch D
sh frame-relay pvc

PVC Statistics for interface Serial0 (Frame Relay DTE)

Active Inactive Deleted Static
Local 1 0 0 0
Switched 0 0 0 0
Unused 0 0 0 0

DLCI = 30, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial0.1

input pkts 1531285 output pkts 1656040 in bytes 697725861
out bytes 159424220 dropped pkts 4 in FECN pkts 0
in BECN pkts 0 out FECN pkts 0 out BECN pkts 0
in DE pkts 1531285 out DE pkts 0
out bcast pkts 61956 out bcast bytes 6252672
Shaping adapts to BECN
pvc create time 12w1d, last time pvc status changed 14:30:13

===================
//Branch E
sh frame-relay pvc

PVC Statistics for interface Serial0 (Frame Relay DTE)

Active Inactive Deleted Static
Local 1 0 0 0
Switched 0 0 0 0
Unused 0 0 0 0

DLCI = 30, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial0.1

input pkts 1453151 output pkts 1647715 in bytes 534304313
out bytes 176339875 dropped pkts 0 in FECN pkts 4
in BECN pkts 0 out FECN pkts 0 out BECN pkts 0
in DE pkts 1453151 out DE pkts 0
out bcast pkts 61998 out bcast bytes 6245972
Shaping adapts to BECN
pvc create time 1w2d, last time pvc status changed 4d09h

====================
//Branch F
sh frame-relay pvc

PVC Statistics for interface Serial0 (Frame Relay DTE)

Active Inactive Deleted Static
Local 1 0 0 0
Switched 0 0 0 0
Unused 0 0 0 0

DLCI = 30, DLCI USAGE = LOCAL, PVC STATUS = ACTIVE, INTERFACE = Serial0.1

input pkts 575764 output pkts 542659 in bytes 389861918
out bytes 72809268 dropped pkts 4 in FECN pkts 0
in BECN pkts 0 out FECN pkts 0 out BECN pkts 0
in DE pkts 575764 out DE pkts 0
out bcast pkts 39307 out bcast bytes 3966486
Shaping adapts to BECN
pvc create time 1d19h, last time pvc status changed 1d19h

 

[Jynxx]

I would actually recommend updating the IOS to a version that supports the interface-congestion parameter...since that is what is actually being monitored to do the adaptive shaping.

We use 12.2.13T4.

Another thing that you can do is specify your CIR to be the 95% value...so it is hard set to never go above that.

One question I do have though...do you have Sprint's basic Frame Relay, or their "Intelligent" Frame Relay (iFR)?

There is a difference. Regardless of which you have, I would definately upgrade code and use adaptive shaping.

 

[ktripp]

Jynxx,
I have a Q for you.
Let's say you have a remote site with a actual circuit of 512 and a true CIR of 256K. Theoretically it should be able to burst up to full port if the providers network isn't congested. Would minCIR be 256K? or 95% of 512K?

Does BE need to be set to zero? From what I've read, it's default is zero.

Thanks for your insight.

KT

 

[Jynxx]

I assume that you do not have Sprint iFR...since you have a CIR...

minCIR in your case would be 256k...since that is the rate at which you are supposedly guaranteed by the carrier.

BE by default is zero, unless otherwise specified. In your case, however, unless you are doing VoIP, I would set your Be to 256k as well and enable adaptive traffic shaping. If you are doing VoIP, leave it at zero. Of course, at that point, you may as well have your port speed reduced to CIR...since it's a waste of money.

What this will do is allow it to "burst" up to port speed, 256k, but if congestion is encountered, to throttle back to your minCIR...or 256k.

 

[OmiDimo]

traffic shaping i think will help you out...
keep in mind that you are always going as fast as your smallest CIR...so if you have 256cir on Z end and a full T1 at the A end, and your A end is sending traffic like mad to the Z end you gonna get a bottleneck.

 

[tetaguila]

traffic shaping only applies if you have a cir ..for example if you have 1024kbps you can divide it by number of routers or users equally...in ur case you have no cir, only port speed.how could you apply the traffic shaper...