[Cerowrt-devel] reproducing network research results with mosh
dave.taht at gmail.com
Fri Mar 22 09:51:50 EDT 2013
On Fri, Mar 22, 2013 at 6:35 AM, <dpreed at reed.com> wrote:
> Just curious... have you run RRUL-style testing of mosh over a congested
> link? Not knowing in what way mosh is "UDP-based" I wonder what happens to
> it when you run a program like, say, "top", over mosh when there is heavy
> file transfer load.
All the time! Given the degree to which I regularly stress out a
network, and crash things, mosh made keeping a dozen or more
connections alive - and interactive - a zillion times less painful.
research paper and more info on how it works here:
Mosh uses a cool protocol that basically does a continuous "diff" of
what should be on the screen. You'll be amused by who wrote the
original idea down... and saddened by how long ago he had done so.
Secondly, given the work I was doing on ECN and classification on top
of red, sfqred, fq_codel, etc over the last 2 years or so, they folded
in ECN support and put in an odd diffserv marking (AF42)
while there isn't a lot of data collected on that that I'm aware of
(keith?) there were two bug reports from the field saying that AF42
traffic had problems on two networks, but thus far the ECN part has
caused no trouble that I know of.
(and keith wasn't aware until a few weeks ago that I'd suggested AF42
because it had an "interesting" bit pattern that would A) put mosh
into the linux wifi VI queue, and B) show the extent to which
providers were mangling the TOS header)
> -----Original Message-----
> From: "Dave Taht" <dave.taht at gmail.com>
> Sent: Friday, March 22, 2013 9:27am
> To: "bloat" <bloat at lists.bufferbloat.net>, "bloat-devel"
> <bloat-devel at lists.bufferbloat.net>, cerowrt-devel at lists.bufferbloat.net
> Subject: [Cerowrt-devel] reproducing network research results with mosh
> I embraced mosh wholeheartedly about a year ago. I've been installing
> it everywhere I go, with much of the same enthusiasm I had for putting
> mosaic on everything back in the early 90s.
> As a ssh replacement for interactive traffic, mosh is of tremendous
> use, particularly in the case where machines are suspended/resumed,
> move access points, or are experiencing wildly variable latency and/or
> long RTTs.
> It's now available across most unix OSes, and often installable via
> the standard package manager.
> Try it!
> It would be awesome if this tech could make it into a generation of
> command line interfaces for embedded gear like routers and switches.
> However at the moment the server code requires C++ and libstd++, which
> makes it overlarge for the lowest end of gear.
> Dave Täht
> Fixing bufferbloat with cerowrt:
> Cerowrt-devel mailing list
> Cerowrt-devel at lists.bufferbloat.net
Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html
More information about the Bloat-devel