Closed tob-00 closed 1 year ago
Hi,
Indeed maybe linked to #435.
Which kernel did you use? Did you compile the latest version of mptcp_v0.96
branch?
Maybe good to try it with different kernel (e.g. mptcp_v0.94
), I may suspect a regression between v0.94 and v0.95. (or maybe older)
Hi, thanks for the fast response! I am using kernel mptcp_v0.95.1
and did not compile on the newer version yet. However, I will try another version.
This version is quite old. I will try to create an mptcp_v0.95.2 in the coming days. Not sure when exactly.
(and maybe a last v0.94 one)
This version is quite old.
Oh really? I thought this is the last stable version as referred to on the homepage on multipath-tcp.org
I thought this is the last stable version as referred to on the homepage on multipath-tcp.org
It is but built 2 years ago. A lot of bug fixes for MPTCP or other parts in the Linux kernel have been added since then.
It is still good just for a quick test but from what we understood, when doing more advanced manipulation, most people are compiling the kernel from one of the stable branches.
It is not difficult to compile a kernel: https://multipath-tcp.org/pmwiki.php/Users/DoItYourself
I just published two new releases: 0.95 and 0.94. Do not hesitate to try those two
@tob-00 I am trying to setup MPTCP in Mininet in Ubuntu and I am really stuck! Would you mind sharing with me which version of Ubuntu, Mininet, and MPTCP you used?
Thank you!
@tob-00 I am trying to setup MPTCP in Mininet in Ubuntu and I am really stuck! Would you mind sharing with me which version of Ubuntu, Mininet, and MPTCP you used?
Thank you!
Hi there! These were the settings I used. I hope it helps!
Hi there, I am running the MPTCP kernel on Ubuntu and I have created a Mininet testbed with both redundant and non redundant topologies. In case I have a non-redundant/single path topology
h1 <--10MBit/s 10ms--> s1 <--10MBit/s 10ms--> s2 <--10MBit/s--> h2
and sending data from h2 to h1, I only get around 2.8MBit/s of throughput and delays up to one second. How can that be? I tested all congestion controls of MPTCP and cubic. For MPTCP with the default and roundrobin scheduler I do not experience anything like that. The throughput is around 9.5MBit/s as expected. I checked wireshark and is says multiple times "TCP windows full". However, I thought the scheduler is only responsible for the subflow selection (of which we only have one) and the congestion control is responsible for the cwnd. I also tried different path managers but this also makes no difference as it shouldn't because there is only one path. My Problem seems to be similar to the issue in this case: https://github.com/multipath-tcp/mptcp/issues/435 . Does anybody have an idea or hint what causes this problem? Thanks a lot in advance. I appreciate any help.