Welcome to this second issue of Multipath News.
Feel free to send suggestions or any feedback to email@example.com or through our online form.
IETF MPTCP mailing list
The IETF mailing list
has been pretty active during April 2017. The first important message
was posted by Christoph Paasch. In this message, he summarises how he has implemented the new handshake proposed for RFC6824bis
. This is an important milestone since this confirms that the new design proposed by Christoph works as expected.
Sayee Kompalli Chakravartula
started several discussion threads
on how to reduce some of the perceived overhead with Multipath TCP. Multipath TCP covers each data with a DSS option, even if there is only one subflow. Sayer argues that the DSS option consumes some space in the header and explores techniques to avoid sending DSS options when there is a single subflow. However, the presence of various types of middleboxes makes this problem more difficult than it could appear at first glance has explained
by Olivier Bonaventure.
Then, most of the bandwidth on the MPTCP mailing list has been consumed by various discussions on a consensus call
that was initiated by the chairs . Various arguments have been raised by people in favour or against pursuing such work within the MPTCP mailing list.
The mptcp-dev mailing list
The mptcp-dev mailing list has been much quieter with only 24 emails exchanged during April 2017
François Finfe has reported
problems with the retransmission of some RST segments that include the MP_FASTCLOSE option. He fixed
this bug later.
The US Department of Energy recently issued its Network2025 report
from its Feb 2016 workshop.
This report provides a roadmap of key networking issues for the short (1-3 yrs), medium (4-6 yrs), and long term (10-12 yrs). It notes that multipath transport protocols, such as MPTCP, could play an important role to support very high speed transmissions over heterogeneous paths.
Feng Zhou et al. use measurements over the Nornet network to analyse in « The Performance Impact of Buffer Sizes for Multi-Path TCP in Internet Setups
» how buffer sizes affect the performance of Multipath TCP. Their main conclusions are that MPTCP provides robust performance and can achieve performance advantages over TCP. Furthermore, buffer requirements remain reasonably small and the performance is independent of the path chosen for the initial subflow.