MPLS Delay

[TheArchitect]

We are installing MPLS links to a number of remote locations. We are seeing substandard speeds to certain locations. In some cases the initial connection to a resource is slow but once the connection is made the speed is fine. We are not seeing any errors on the links. What else can I look at to help troubleshoot this? If its helpful, the provider is using Juniper gear on their side. We are using a combination of 3800 and 2800 series routers.

[swaroop.potdar]

Can u ping with an MTU of 1500 with the DF bit set from remote locations to the resource servers. See if this goes successful.

Apart form this MPLS would not play much more of a role in your end to end connection.

Do test this and put in more observations so you may get desired reply.

HTH-Cheers,

Swaroop

[TheArchitect]

Ping -s 1500 -M do x.x.x.x fails from FC3 machine.(-M do shuts off fragmentation) I would expect that though as 1508 actual size of the packet exceeds the MTU of the interface. Ping -s 1400 -M do x.x.x.x is successful.

The T1 interfaces have MTU set @ default 1500. The 3mb DS3 links are @ default 4470.

Ping -R from Fedora returns a consistant anomalous hop on the return path that is not seen in a traceroute. I cannot explain this. The -R option manpage docs talk about this flag returning cache results. The test results have been consistant over several days. I would expect bad information to get dropped fairly quickly rom a cache. I have asked the provider to investigate.

All routing inside the MPLS cloud is handled by the provider. Our routes into the cloud are static default routes.

[swaroop.potdar]

To overcome the MTU problem there are two things which can be done.

1) Enable TCP MTU discovery on the all routers. and Provider Edge routers also to support the same.

2) Or The provider should be capable of supporting Jumbo Frames without fragmenting.

HTH-Cheers,

Swaroop

PS: You may want to refer to this past link for more info.

" MPLS: MTU Ethernet MPLS Network "

[mheusinger]

Hi,

performance problems can have many reasons. Among them are MTU related problems mentioned by swaroop but also a lack of resources can contribute. The latter could arise from overload situations in the SP network or overloading your network, especially the access lines. QoS might help to solve those issues by prioritizing important traffic.

Troubleshooting involves to localize the issues. One thing to look into would be "ip sla" (www.cisco.com/go/ipsla). In brief, your routers could be setup to monitor packet drops, jitter and delay across the SP network. IP SLA options are to access HTTP, monitor TCP connect times, jitter and delay for UDP and so on.

Hope this helps! Please rate all posts.

Regards, Martin

[colbedi]

Hi

We have faced similar problems. Our setup is that we have 7206 routers as PE and are using l2tpv3 for far end connections. Now one fine when the connectivity to the CE breaks and is restored , the end to end ping is coming but application stop.

regards

Aasheesh GUpta

[chintu_punjabi]

Pls make sure that ur MTU size is same on ur PE >> PE / PE >> P links.

HTH

Chintu