MLPPP configuration on CE and PE

[mohammedmahmoud]

Hi All,

I configured the PE and the CE as follows, but i got a lot of packet loss and a very high delay, any comments ?

CE:

!

interface Loopback13

ip address 13.13.13.2 255.255.255.252

!

interface Virtual-Template1

bandwidth 2048

ip unnumbered Loopback13

ppp multilink

ppp multilink fragment-delay 10

ppp multilink interleave

!

interface Serial0

no ip address

encapsulation frame-relay IETF

no fair-queue

frame-relay traffic-shaping

frame-relay lmi-type cisco

!

interface Serial0.1 point-to-point

frame-relay class mlppptest

frame-relay interface-dlci 101 ppp Virtual-Template1

!

interface Serial1

no ip address

encapsulation frame-relay IETF

no fair-queue

frame-relay traffic-shaping

frame-relay lmi-type cisco

!

interface Serial1.1 point-to-point

frame-relay class mlppptest

frame-relay interface-dlci 102 ppp Virtual-Template1

!

!

map-class frame-relay mlppptest

frame-relay adaptive-shaping becn

frame-relay cir 2048000

frame-relay bc 20480

frame-relay be 0

!

PE: (ATM Router)

!

interface Loopback13

ip address 13.13.13.1 255.255.255.252

!

!

interface Virtual-Template1

bandwidth 2048

ip unnumbered Loopback13

ppp multilink

ppp multilink fragment delay 14

ppp multilink interleave

!

interface Switch1

no ip address

no atm ilmi-keepalive

no rpm-sar-auto-recovery

rpm-auto-cbclk-change

!

interface Switch1.101 point-to-point

description MLPPP Test 1st Link

pvc 0/101

abr 2663 2663

encapsulation aal5autoppp Virtual-Template1

!

!

interface Switch1.102 point-to-point

description MLPPP Test 2nd Link

pvc 0/102

abr 2663 2663

encapsulation aal5autoppp Virtual-Template1

!

!

Thanks in advance.

[mohammedmahmoud]

After changing the CE (it was 1721 and i replaced it with 1751V) --> The delay and packet drop issue was solved but the links utilization doesn't exceed 1.5M.

Any suggestions ?

Thanks in advance.

[pkhatri]

Hi,

One thing I would suggest is to remove the 'bandwidth' statement from the virtual-template interface definition. When using MLPPP, you generally want your bandwidth to be determined dynamically depending on the number of active interfaces in the bundle.

Hope that helps - pls rate the post if it does.

Paresh

[mohammedmahmoud]

Paresh,

Thanks alot, but i've tried you suggestion and it didn't work --> The links still saturate @ 1.5M.

Any other suggestions ?

[pkhatri]

So you are getting 1.5Mbps per link, right ? And not 1.5Mbps of aggregate traffic ?

One other thing you might want to try is remove ' frame-relay adaptive-shaping becn' from your FR map-classes. That will ensure that you are always shaping out at 2048k.

Could you post the output of 'sh frame-realy pvc 101' and 'sh frame-realy pvc 102' from your CE ?

Paresh

[mheusinger]

Hello,

can you increase the frame-relay fragment size 14 from the config for testing purposes? This is rather tough on CPU and doesn´t work without problems. Once you found it solves the problem then try something like 20 ms or 30 ms, depending on what a good compromise is between jitter for voip and performance of the interface. I once had this problem with 2610XM. They were also causing a much higher jitter than the configuration value suggested. Increasing the fragment size lowered (!) the jitter to acceptable values.

Hope this helps! Please rate all posts.

Regards, Martin

[mohammedmahmoud]

All, the problem was solved by removing any fragmentation (disabling fragmentation) --> We found in one of the CCO documents that for links with BW higher than half the T1 bw there should be no fragmentation.

http://www.cisco.com/en/US/partner/tech/tk652/tk698/technologies_tech_note09186a0080094660.shtml#stpri

Thanks for your help.

[mohammedmahmoud]

All,

I am sorry but i still have a problem:

Everything looks OK when using a traffic generator (TfGen), but when using FTP it never peaks, is there any known problem for FTP with MLPPP.

I even tried per-packet load balancing with CEF, but FTP also intorduces problems, is there any known problems for FTP, with either CEF per-packet or MLPPP.

Thanks in advance.

[pkhatri]

Hi,

FTP users TCP, with responds to congestion and dropped packets by dropping its transmission rate. Therefore, when using a single FTP connection, you will find that the effective throughput will alwyas be less than the line-rate. That is just how TCP works...

IF you try and run multiple FTP sessions, you will find that you start approaching the line rate. It's a good idea to implement some form of RED though.

I would not recommend that you disable CEF, as your performance will suffer.

Hope that helps - pls rate the post if it does.

Paresh