[Starlink] congestion control info off the datapath

[Starlink] congestion control info off the datapath

[Dave Taht]

One of our newer members is one of the authors of this:

https://people.cs.rutgers.edu/~sn624/papers/ccp-sigcomm18.pdf

I will reserve comment for now. I don't grok the intersection with ebpf as yet.

Please? newer folk here, please briefly introduce yourselves and your
interests in starlink's stuff?
I very much approve of lurkers - even aliases! for those that cannot
talk due to various NDAs, etc, but...

For those here that don't know me already, well, I'm this renegade
that has long lived
outside of the vogon-industrial-military-academic complex with more
key open source
networking contributions to the world than most. These days I'm best
known for the bufferbloat
project and multiple congestion control algorithms/packet schedulers and AQMs

( Please put a copy of https://bufferbloat-and-beyond.net/ in your
reading queue)

but I have a long history going back to 1982 of kicking the tires of
the internet,
creating a RUD, and then fixing it.

My own personal mission (as per the podcast) is to merely see starlink
get sch_cake up on outbound
on the dishy but long term I care a lot about just about anything
networked and in
space from here out to the edges fo the solar system. I'm really low
on spare time
at the moment personally, but if newer folk here would like to
introduce themselves
and point a link at a key paper of theirs or at their work in
progress, perhaps the
conversation will get rolling a bit more. We're still looking for more
folkable  to do
our measurement suite in particular.

(we also need to put up the mailing list archive so far for new folk.
Am short an
 mailman expert for that if anyone has those chops?)

-- 
Fixing Starlink's Latencies: https://www.youtube.com/watch?v=c9gLo6Xrwgw

Dave Täht CTO, TekLibre, LLC

Re: [Starlink] congestion control info off the datapath

[Ulrich Speidel]

[-- Attachment #1: Type: text/plain, Size: 2787 bytes --]

I'm Ulrich and I'm at the University of Auckland, New Zealand. Broader 
interest is in novel technologies for getting (better or any at all) 
Internet to the underserved of this world. In our backyard that means - 
mostly but not only tropical - Pacific Islands with too few folks on 
them to make a business case for a submarine fibre connection. I've so 
far mostly concentrated on narrowband GEO and MEO pipes to small island 
ISPs or ISP-like entities, and you've linked to our home page 
(https://sde.blogs.auckland.ac.nz/ 
<https://sde.blogs.auckland.ac.nz>) 
the other day - much appreciated.

Interest in Starlink: It's one of the developments that might help in 
this effort, although the current orbital inclinations (53 degrees) 
still leave large gaps around the tropical regions. No surprise all 
those gushing beta user reports come from people north / south of 40 
degrees latitude, where satellite density is highest. For the tropical 
Pacific, remoteness will be another showstopper while Starlink uses 
standard bent pipe links - you really need to be within a couple of 
hundred miles of a gateway teleport of this to work. Many islands we 
worry about are many hundreds of miles from the next connected piece of 
land, let alone a Starlink teleport. So once inter-satellite routing 
will be happening (how? when? where?), we may be debating bufferbloat 
all over again. I'm also intrigued how the global Starlink system 
capacity (rumoured to be 23.6 Tb/s for the commercial start) will 
suffice to service the billions of underconnected in the world when much 
of that capacity will be over water at any time (where it's not needed). 
I'm thinking here in particular that this number is awfully close to the 
current connected capacity between Australia/NZ and North America, and 
there's only around 25 million of us here who, unlike Starlink 
customers, have CDN server farms between ourselves and our international 
links. So I have heaps of questions, which have recently appeared in an 
APNIC blog:

https://blog.apnic.net/2021/05/20/everything-you-wanted-to-know-about-leo-satellites-part-1-the-basics/ 


On 25/07/2021 5:02 am, Dave Taht wrote:
>
> Please? newer folk here, please briefly introduce yourselves and your
> interests in starlink's stuff?
> I very much approve of lurkers - even aliases! for those that cannot
> talk due to various NDAs, etc, but...
-- 

****************************************************************
Dr. Ulrich Speidel

School of Computer Science

Room 303S.594 (City Campus)
Ph: (+64-9)-373-7599 ext. 85282

The University of Auckland
ulrich@cs.auckland.ac.nz
http://www.cs.auckland.ac.nz/~ulrich/
****************************************************************




[-- Attachment #2: Type: text/html, Size: 3736 bytes --]

Re: [Starlink] congestion control info off the datapath

[Andrew Crane]

[-- Attachment #1: Type: text/plain, Size: 1148 bytes --]

< Please? newer folk here, please briefly introduce yourselves and your
< interests in starlink's stuff?

My story.
I'd moved from a suburban area, which had decent DOCSIS service, to a
remote rural area, where the choice was between woeful DSL service... or
nothing.
After a few weeks of pain, neighbors alerted me to a home router that
supposedly made life bearable. I'd encountered a lot of worthless IT snake
oil in my job & was skeptical. But I did some reading about bufferbloat,
threw skepticism to the wind, and bought the router.
It tuned itself to the connection and ended up making things workable. I'm
not a network engineer  - my specialty was DNS/DHCP/IP address management -
but nerded out around the subject of latency under load. I ended up putting
together an OpenWRT-based home router.
When Starlink became operational, bufferbloat was the first thing I was
curious about, especially given the "back-to-first-principles" engineering
that Starlink aspires to.
~ Andrew


> _______________________________________________
> Starlink mailing list
> Starlink@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/starlink
>

[-- Attachment #2: Type: text/html, Size: 1806 bytes --]

Re: [Starlink] congestion control info off the datapath

[Jared Mauch]

On Sat, Jul 24, 2021 at 10:02:43AM -0700, Dave Taht wrote:
> Please? newer folk here, please briefly introduce yourselves and your
> interests in starlink's stuff?
> I very much approve of lurkers - even aliases! for those that cannot
> talk due to various NDAs, etc, but...

Sure, I'm Jared Mauch, I work at Akamai ($dayjob) and my night job is
running a small FTTH business.  I have starlink at a family cabin that
is off-grid about 1k miles away from home.  Because of $night_job I need
to have periodic access to the network remotely.

In my additional spare time I enjoy running my chainsaw, or going rock
climbing both in-gym and outdoors.  I spend a lot of time paying
attention to things like broadband in rural areas and how it can be
expanded.  This includes helping other providers that are on a smilar
path to build FTTH in rural areas.

	- Jared

Re: [Starlink] congestion control info off the datapath

[Michael Richardson]

[-- Attachment #1: Type: text/plain, Size: 971 bytes --]


< Please? newer folk here, please briefly introduce yourselves and your
< interests in starlink's stuff?

I live in downtown Ottawa, and I have VDSL2 (because the other options have
no IPv6, and are massively bufferbloated).

But, my relatives choose to live in partially off the grid (at the edge of
the grid?), eat "organic" and drive hundreds of kilometers a week for work.
(I walk to the store, am mostly car-free, and think vegetables that consume
huge amounts of CO2 to get to me can never be more healthy for the planet).

I care about Starlink because it might let my relatives stop driving,
and because it likely will give the Canadian bufferbloated duopoly a serious challenge.

But, it's gotta be real terminal-to-terminal, or it will just result in more
"bent fiber" pervasive attacks .

--
Michael Richardson <mcr+IETF@sandelman.ca>   . o O ( IPv6 IøT consulting )
           Sandelman Software Works Inc, Ottawa and Worldwide





[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

Re: [Starlink] congestion control info off the datapath

[Srinivas Narayana]

On Sat, Jul 24, 2021 at 10:02:43AM -0700, Dave Taht wrote:
> 
> Please? newer folk here, please briefly introduce yourselves and your
> interests in starlink's stuff?
> I very much approve of lurkers - even aliases! for those that cannot
> talk due to various NDAs, etc, but...

Hi everyone!

I'm an academic at Rutgers University, New Jersey, USA. I've
been interested in programmable networking research, defined
in the broadest possible terms. I've done research on
congestion control/transport before, and have background in
both endpoint and in-network (router) scheduling. I've also
been following Dave's (and others') cool work on router AQMs
and debloating.

Through the pandemic and after spending time in India (with
some pretty bloated data networks, especially cellular),
I've gained a visceral appreciation for low latency
networking. These days I'm interested in transport and
scheduling technologies, as well as novel options for
connectivity (starlink included), that can help us end users
to experience a better Internet in its all realtime glory.


> One of our newer members is one of the authors of this:
> 
> https://people.cs.rutgers.edu/~sn624/papers/ccp-sigcomm18.pdf
> 
> I will reserve comment for now. I don't grok the intersection with ebpf as yet.

Thanks for posting a pointer to this :) I'm happy to answer
any questions about the work.

-- 
Srinivas
An adequate bootstrap is a contradiction in terms. 

Re: [Starlink] congestion control info off the datapath

[Nick Buraglio]

[-- Attachment #1: Type: text/plain, Size: 3067 bytes --]

Brief introduction from me:
I'm Nick Buraglio.I've been building service provider networks in one way
or another since the 90s. My day job is on the planning and architecture
team for the Energy Sciences Network  (es.net). I've been doing HPC and LFN
networks for about 20 years now, having worked on almost every aspect from
security policy and implementation to fiber planning and optical wave
multiplexing, and most things in between (Layer 1 to Layer 8). My interest
in StarLink is mostly personal, as I have long been a strong proponent of
providing useful resources to underserved areas, but also from an
academic perspective in dealing with the technical pieces. Traffic
engineering, IPv6 proliferation, as well as performance testing are a big
part of my drivers and personal interests. Sometimes I do a podcast called
modem.show that talks about technical edge cases and deeper dive
take-aparts.

nb


On Sat, Jul 24, 2021 at 12:02 PM Dave Taht <dave.taht@gmail.com> wrote:

> One of our newer members is one of the authors of this:
>
> https://people.cs.rutgers.edu/~sn624/papers/ccp-sigcomm18.pdf
>
> I will reserve comment for now. I don't grok the intersection with ebpf as
> yet.
>
> Please? newer folk here, please briefly introduce yourselves and your
> interests in starlink's stuff?
> I very much approve of lurkers - even aliases! for those that cannot
> talk due to various NDAs, etc, but...
>
> For those here that don't know me already, well, I'm this renegade
> that has long lived
> outside of the vogon-industrial-military-academic complex with more
> key open source
> networking contributions to the world than most. These days I'm best
> known for the bufferbloat
> project and multiple congestion control algorithms/packet schedulers and
> AQMs
>
> ( Please put a copy of https://bufferbloat-and-beyond.net/ in your
> reading queue)
>
> but I have a long history going back to 1982 of kicking the tires of
> the internet,
> creating a RUD, and then fixing it.
>
> My own personal mission (as per the podcast) is to merely see starlink
> get sch_cake up on outbound
> on the dishy but long term I care a lot about just about anything
> networked and in
> space from here out to the edges fo the solar system. I'm really low
> on spare time
> at the moment personally, but if newer folk here would like to
> introduce themselves
> and point a link at a key paper of theirs or at their work in
> progress, perhaps the
> conversation will get rolling a bit more. We're still looking for more
> folkable  to do
> our measurement suite in particular.
>
> (we also need to put up the mailing list archive so far for new folk.
> Am short an
>  mailman expert for that if anyone has those chops?)
>
> --
> Fixing Starlink's Latencies: https://www.youtube.com/watch?v=c9gLo6Xrwgw
>
> Dave Täht CTO, TekLibre, LLC
> _______________________________________________
> Starlink mailing list
> Starlink@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/starlink
>

[-- Attachment #2: Type: text/html, Size: 4004 bytes --]