strange mpls ?

[xuetao han]

We started switching to a new mpls provider.

First step, two 200M links were commissioned, each terminated in one of my DC; two DC also have dark fibre as backbone.

There is no dynamic ce/pe routing as i am going to build my Flexvpn.

To test mpls link throughput,  a plain GRE tunnel was created on two cpe over mpls, RTT is 30ms

 Test was done by iperf. here are the figures:

   tcp window size             calculated throughput         recorded throughput

        64KB                                        17M    bps                                  9M bps

       256K                                          68M                                            17M

       512K                                          136M                                          17M

but iperf with UDP can achieve 160Mbps.

to prove my gears are capable, another gre tunnel is created between these exact two cpe, via my backbone fibre. When testing with iperf. it reached 400Mbps with 64K window and 800M with 256K window.

I asked new service provider to do  rfc 6349 testing but was told they had only rfc 2544 tools.

Any idea ?

[Vasilii Mikhailovskii]

Hello.

I guess your provider could mean 200M as 200*1000*1000 bps, while your 160Mbps is about 167 *10^6 bps.

If you are running GigEthernet, then overhead could be 38 (L1+L2)+20 (L3)+20(L4) per 1460 bytes payload => 167*10^6 *1538 / 1460 = 176.7 *10^6.

I'm not sure how were you running iperf and if you had router (with shaping feature) on the path. I would suggest to ask your ISP to provide test per rfc 2544.