Starlink has bufferbloat. Bad.
 help / color / mirror / Atom feed
From: Nathan Owens <nathan@nathan.io>
To: Ulrich Speidel <ulrich@cs.auckland.ac.nz>
Cc: starlink@lists.bufferbloat.net
Subject: Re: [Starlink] Starlink tidbits from NANOG
Date: Thu, 4 Nov 2021 18:05:17 -0700	[thread overview]
Message-ID: <CALjsLJu8YiC0sKRX6VjOyoiHd4URDRLd_P6OW4nvtX5vBjuGmA@mail.gmail.com> (raw)
In-Reply-To: <4ea38008-0e62-1cad-165a-a8aad232ebd6@cs.auckland.ac.nz>

[-- Attachment #1: Type: text/plain, Size: 4954 bytes --]

> They are using waves back to regional DCs now, but will be moving to
> dark fiber over the next year or two
If that means "radio" waves, then this goes a long way to explaining why
there's already limited capacity even near the US-Canada border.

Waves in this case generally refers to 10G/100G leased optical circuit
capacity.

On Thu, Nov 4, 2021 at 5:35 PM Ulrich Speidel <ulrich@cs.auckland.ac.nz>
wrote:

> Thanks for that Darrell - that's really interesting! A few comments on
> that front:
>
> On 5/11/2021 4:26 am, Darrell Budic wrote:
> > I was at NANOG in Minneapolis, and got a chance to ask a couple
> > question of a Starlink Network Engineer who’s attending. I was already
> > talking to him about Starlink’s network efforts (see below) but it was
> > nice to meet in person. Don’t quote me on any of this, but here’s a
> > few tidbits this list may appreciate:
> >
> > - Starlink is expanding their own network operations, and is
> > connecting to more IXPs. They were already on SIX in Seattle, have
> > connected to DECIX NY, and are in the process of connecting to ChIX in
> > Chicago. As I run ChIX, I had a good excuse to talk to them about
> > other things. :) IXPs and their own networks are in the works for
> > Europe and other areas as well.
> Makes sense.
> > - They have been obtaining more v4 addresses, but I don’t know if they
> > have enough to not do CGNAT. I don't think they do yet, but it seems
> > like it may be a long term target.
> > - v6 is deliberately not fully functional, but they know some of use
> > are using it and it will eventually be fully activated. May be waiting
> > on the regional connectivity, so will be intersting to see if changes
> > for some areas and not others as they roll it out.
>
> So I guess we need to distinguish between:
>
> - IPv4 addresses for any CGNAT they might run
> - IPv4 addresses as static addresses for (some of?) their customers
> - IPv6 addresses as customer addresses
> - IPv6 addresses to support geographic routing as discussed in earlier
> posts (subnet maps to cell / satellite)
>
> There are quite a number of feasible configurations in this. E.g., they
> could be running a CGNAT setup with a v4 pool on the Internet side, use
> v6 to tunnel route from there to the satellite the end customer connects
> to, and then map that customer back to a (private) IPv4 address in a NAT
> on the satellite. One aspect that hasn't really been mentioned much here
> is that of PDU size on the link between end customer and satellite.
> Keeping Dishy and its successors small and cheap creates an incentive to
> operate at marginal SNR, and this favours smaller PDUs over larger ones
> as the probability of PDU checksum errors increases with PDU size. But
> having lots of small PDUs means having lots of headers, and as IPv4
> addresses are leaner than IPv6 ones, this saves bandwidth here. Probably
> not a biggie though.
> > - New ground stations with more capacity are coming (and will be
> > upgrades).
> Any word on where? At the moment, most of the world can see Starlink
> satellites, but most Starlink satellites can't see a ground station.
> > They are using waves back to regional DCs now, but will be moving to
> > dark fiber over the next year or two
> If that means "radio" waves, then this goes a long way to explaining why
> there's already limited capacity even near the US-Canada border.
> > - the new satellites have more than 2 lasers, and there is enough
> > capacity on them to do routing. no details on how or what protocols, alas
> Any word on when we can expect to see routing in action?
> > - new birds also have 2-3x more ku bandwidth than first gen
> Hm. Sounds cool, but with 3 billion or so underserved on the planet &
> typical annual growth rates, that's still just a drop in the bucket.
> > - new dishes are in the works, v4 coming with lower power use, more
> > capacity, not round any more
> Trayee? Squary? Just joking ;-)
> > - larger dishes coming for commercial apps
> That's good news, as this will allow Starlink to be used in places where
> direct-to-site crashes into regulatory hurdles. If we can get the big
> CDN providers to come up with small (virtual?) appliances that can be
> put at the remote end of such links by local ISPs, then that'll also
> help to preserve space segment capacity.
>
> --
> ****************************************************************
> 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/
> ****************************************************************
>
>
>
> _______________________________________________
> Starlink mailing list
> Starlink@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/starlink
>

[-- Attachment #2: Type: text/html, Size: 6587 bytes --]

  reply	other threads:[~2021-11-05  1:05 UTC|newest]

Thread overview: 22+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-04 15:26 Darrell Budic
2021-11-04 16:21 ` Dave Taht
2021-11-04 16:46 ` Dave Taht
2021-11-04 17:11   ` Dave Taht
2021-11-05  1:30     ` Darrell Budic
2021-11-05  1:41   ` Darrell Budic
2021-11-05  1:46   ` Darrell Budic
2021-11-04 18:16 ` Michael Richardson
2021-11-04 18:25   ` Inemesit Affia
2021-11-04 18:29   ` Dave Taht
2021-11-04 19:10     ` Michael Richardson
2021-11-05  0:34 ` Ulrich Speidel
2021-11-05  1:05   ` Nathan Owens [this message]
2021-11-05  1:18     ` Darrell Budic
2021-11-05 22:24       ` Ulrich Speidel
2021-11-05 15:00     ` [Starlink] data sovereignty Michael Richardson
2021-11-05 15:07       ` Spencer Sevilla
2021-11-05 17:36       ` Daniel AJ Sokolov
2021-11-05 18:01       ` Daniel AJ Sokolov
2021-11-05 22:12     ` [Starlink] Starlink tidbits from NANOG Ulrich Speidel
2021-11-05  1:27   ` Darrell Budic
2021-11-05 15:03     ` Michael Richardson

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: https://lists.bufferbloat.net/postorius/lists/starlink.lists.bufferbloat.net/

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CALjsLJu8YiC0sKRX6VjOyoiHd4URDRLd_P6OW4nvtX5vBjuGmA@mail.gmail.com \
    --to=nathan@nathan.io \
    --cc=starlink@lists.bufferbloat.net \
    --cc=ulrich@cs.auckland.ac.nz \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox