How to prove MPLS network losses to Telco

[Amol Akadkar]

We are using MPLS network and facing slowness in application access, some time all application access stop because of congestion or losses.

We logged complaint with Telco and they refused that there is no losses on the link. they used Ping flood and provided results to us.

I am sure, Ping is not the only or exact way to determine the losses on the link.

For testing we used 'iperf' utility and it shows losses sometimes DC to branch or Branch to DC. These losses are around 15 to 20% consistent.

My question is, 'iperf' toll results are how much reliable?

Is there a way or tool available, to prove Telco that there are losses on the link?

Regards,

Amol

[Akash Agrawal]

Hi Amol,

How is your topology? you have taken just physical link from Telco or using their MPLS infrastructure to connect DC to branch. 

You can configure ip sla to monitor layer 3 reachability. when there will be any drop in the path, ip sla will go down. To check in which portion packet drop is happening, you can configure multiple ip sla to different point in the path. Suppose your path is R1--R2--R3--R4--R5 and drops are observed between R1 and R5. You can set ip sla from R1 to R2,R3,R4. during issue if ip sla to R2,R3 is up and R4-ipsla flaps that means issue is between R3 and R4.

Packet drop could be due to

- link layer1 issue (CRC will increase on interfaces in the path)

- Congestion (output drops will increase on interface in the path)

- Unstable routing in the network (route to destination will get refreshed)

http://www.cisco.com/c/en/us/td/docs/ios/12_4/ip_sla/configuration/guide/hsla_c/hspthjit.html#wp1051321

Regards,

Akash

[ashirkar]

Hello Amol,

I feel ,you should check result with certain MTU ,sometimes telco doesn't have end to end same mtu which affects some applications but normal ping works fine,i guess its called  mss (maximum segment size).As suggested by akash ,you can also check delay and drops for UDP or icmp packet using IP SLA.

Regards,

Ashish 

[Tagir Temirgaliyev]

it is easy to check

ping from end to end with size 1500 df

if ping is ok so mtu 1500

if no pings so reduce size and ping again