Forgot your password?
typodupeerror
ISS Mars NASA Networking Robotics Space The Internet Hardware News

NASA DTN Protocol: How Interplanetary Internet Works 109

Posted by samzenpus
from the in-space-nobody-can-hear-you-click dept.
First time accepted submitter GinaSmith888 writes "This is a deep dive in the BP protocol Vint Cerf developed that is the heart of NASA's Delay-Tolerant Networking, better known as DTN. From the article: 'The big difference between BP and IP is that, while IP assumes a more or less smooth pathway for packets going from start to end point, BP allows for disconnections, glitches and other problems you see commonly in deep space, Younes said. Basically, a BP network — the one that will the Interplanetary Internet possible — moves data packets in bursts from node to node, so that it can check when the next node is available or up.'"
This discussion has been archived. No new comments can be posted.

NASA DTN Protocol: How Interplanetary Internet Works

Comments Filter:
  • First post (Score:5, Funny)

    by JonWan (456212) on Sunday November 11, 2012 @12:16PM (#41949877)

    The main problem is the long delay at light speed.

    • by JonWan (456212) on Sunday November 11, 2012 @12:19PM (#41949901)

      Holy shit! that just blows the crap out of that joke. All these years on Slashdot and the one time I try to do a first post joke I really get first post.

      • Re: (Score:3, Funny)

        by SteveFoerster (136027)

        Were you measuring light speed in American units instead of metric ones?

        • by JonWan (456212)

          Nah, I'm in Texas we use furlongs per fortnight.

          • by stfvon007 (632997)

            According to google:
            the speed of light = 1802617500000 furlongs per fortnight
            Though it may be less for you, since everything is bigger in texas (bigger furlongs = lower speed of light in fpf). Ive been wondering though, since everything is bigger in texas, does that mean everything is redshifted due to the wavelengths being longer?

        • There's no need to create confusion about that; how about being DTN compatible?

          it’s a protocol called Delay-Tolerant Networking, better known as DTN.

          NASA’s experimental Disruption Tolerant Networking (DTN) protocol

      • by RockDoctor (15477)
        Life's a beach and then you fry!

        Trying to drag the quality of First Post comments up is always going to be a futile game. But thanks for trying.

    • Having spent the last 3 weeks in the Galapagos, I can assure you there are long delays and dropouts even in tourist spots here at home(earth)... I do wish someone would implement some better protocol for those locations where speed and reliability aren't up to a reasonable standard. It would definitely help with satellite based ISPs in out of the way locations.. Having a system that would optimize the transaction(automatically send images, js, css despite not requesting it yet..) and automatically re-reques

      • by delt0r (999393)
        Its probably easier and cheaper to just get/pay/invest in a more reliable network. Don't forget all the software that would need to be DTN aware.
        • When all restaurants, hotels and internet cafes tested can't provide sufficiente quality to maintain a 2 second voice only skype call, it might be that just "upgrading your plan" might not be enough...
          It doesn't always make sense to buy, and travel with, to third world countries, state-of-the-art satellite communication devices... (They are using non-state-of-the-art satellite and microwave systems and they aren't doing very well...)
          They are hoping to get fiber-optic to the islands, but you can't just insta

      • by Muad'Dave (255648)

        In the old days, we had exactly what you describe with regard to protocol optimization. A company I worked for used VSAT [wikipedia.org] comms to all their stores, and the VSAT hardware would spoof the IP and X-25 protocols on either earthbound end. What went over the sat link was a very optimized protocol tailored for the relatively long double hop delay.

  • TLDR X.25 with big buffers

  • by yincrash (854885) on Sunday November 11, 2012 @12:18PM (#41949891)
    :( :( :(
  • by DamageLabs (980310) on Sunday November 11, 2012 @12:26PM (#41949941) Homepage

    I thought something like this already existed. And it worked pretty well at the time.

    http://en.wikipedia.org/wiki/UUCP [wikipedia.org]

  • by TheDarAve (513675) on Sunday November 11, 2012 @12:29PM (#41949965)

    The only problem with the BP protocol is the data mining rigs that burst and spread raw SQL queries all over the coast of Amazon.com and then wonder why people are pissed that they can't buy or sell from that site until its cleaned up!

  • ...with the tech lower in the piece"

    Except, they never do.

  • by AllanNienhuis (2772209) on Sunday November 11, 2012 @12:55PM (#41950115) Homepage
    Reminds me of the problems the old FidoNet had to deal with - nodes not being available, or only available for short times, poor quality connections, low speed, etc. It worked remarkably well for all of those conditions I thought :)
  • by a_hanso (1891616) on Sunday November 11, 2012 @01:00PM (#41950149) Journal
    While on the subject, when are we going to establish repeater stations around the solar system so that space probes don't need massive transceivers and line-of-sight to communicate with the Earth?
    • by Zocalo (252965)
      Sounds like a good idea, but I suspect there would be a problem with signal power and/or fuel. A deep space relay would need to be able to either receive signals from any direction, at any time, or to point a directional antennae towards a given point in the sky whenever a signal is due. The former approach means that it becomes a lot more difficult to pickup a faint signal and the latter is almost certainly going to require fuel in order to either keep the satellite stable as antennae move around or to r
      • by Nyeerrmm (940927) on Sunday November 11, 2012 @02:15PM (#41950681)

        Actually, the DSN model would work fine for a manned mission to Mars. You're never actually behind the Sun (well it might be possible, but it would be for less than a day). There is an issue where the Sun-Earth-Probe angle drops down to around 3-degrees (so, close to behind the sun), because of radio interference from the sun, but thats about a week long period that you could probably get away with. The biggest cause of comm issues at Mars is Mars itself. Fortunately, all orbiters have an Electra package that allows them to act as relays for each other and for surface assets.

        Relay systems are actually more useful in the Earth-moon system at this point. A Lagrange point relay would be important for a far-side base on the moon, or a lander on that side. Earth orbit is where the biggest need for relays is, because the Earth is always in the way for LEO assets. Thats why we have TDRSS.

        The biggest issue right now is simply the load on the DSN. Its underfunded and its hard to get enough time on it.

    • by Zoxed (676559)

      Would this count as a start ? Artemis [wikipedia.org] (SKDR S/Ka band Data Relay).

    • by Thuktun (221615)

      While on the subject, when are we going to establish repeater stations around the solar system so that space probes don't need massive transceivers and line-of-sight to communicate with the Earth?

      IIRC, isn't that what the Mars orbiters are effectively doing for all the rover missions?

  • Isn't IP already delay tolerant? I remember in the IPoAC trial for obvious reasons there were huge delays, but it still worked.

    • by isorox (205688)

      Isn't IP already delay tolerant? I remember in the IPoAC trial for obvious reasons there were huge delays, but it still worked.

      Ip over air canada? Yes certainly delays and rerouting.

      Ip over avian carrier can cope with high delays and dramatic jitter, re-ordering and packet loss. I'm not sure udp/tcp can cope though, and Ip itself can't cope very well with a situation where the route only partially exists (say your orbiter acts as a router but is on the other side of the planet to your target lander, and needs to store the packets for a few hours)

      • Isn't IP already delay tolerant? I remember in the IPoAC trial for obvious reasons there were huge delays, but it still worked.

        Ip over air canada? Yes certainly delays and rerouting.

        Ip over avian carrier can cope with high delays and dramatic jitter, re-ordering and packet loss. I'm not sure udp/tcp can cope though, and Ip itself can't cope very well with a situation where the route only partially exists (say your orbiter acts as a router but is on the other side of the planet to your target lander, and needs to store the packets for a few hours)

        Actually he was talking about IPoverAnonymousCoward.

        Or did you forget where you were?

    • by Zoxed (676559)

      I am running a test: still waiting on a result:

      ping -W 118000 voyager1.nasa.gov

    • Isn't IP already delay tolerant? I remember in the IPoAC trial for obvious reasons there were huge delays, but it still worked.

      Huge, yes - but not astronomically huge. The main failing of TCP/IP in extra-planetary usage is that RTT/delay measured in minutes smacks up against many-and-various timeouts in TCP. Effectively TCP thinks RTT of 16 minutes (eg to the sun and back) is actually 100% packet loss, because TCP gave up waiting (timeout) ages ago.

      • by danhuby (759002)

        Huge, yes - but not astronomically huge.

        The delays were longer than 16 minutes... over an hour in most cases. There's the printing of the data, strapping it to the pigeon, scanning it back in, it all adds time.

        I'm not sure if they increased timeouts to cope with the problem.

  • And now the question is: when will Linux support it?

    • by Anonymous Coward

      ION is probably the most popular open source implementation of DTN, and was developed on Linux machines..

    • Re:Linux (Score:4, Informative)

      by John Hasler (414242) on Sunday November 11, 2012 @04:32PM (#41951577) Homepage

      > when will Linux support it?

      Package: ion
      Version: 3.0.1~dfsg1-1
      Installed-Size: 2618
      Maintainer: Leo Iannacone
      Architecture: amd64
      Depends: libion0 (= 3.0.1~dfsg1-1), libc6 (>= 2.7), libexpat1 (>= 2.0.1)
      Suggests: ion-doc
      Description-en: NASA implementation of Delay-Tolerant Networking (DTN)
        Interplanetary Overlay Network (ION) software distribution
        is an implementation of Delay-Tolerant Networking (DTN)
        architecture as described in Internet RFC 4838.
        .
        This is a suite of communication protocol implementations designed
        to support mission operation communications across an end-to-end
        interplanetary network, which might include on-board (flight) subnets,
        in-situ planetary or lunar networks, proximity links,
        deep space links, and terrestrial internets.
        .
        Included in the ION software distribution are the following packages:
          * ici (interplanetary communication infrastructure) a set of libraries
              that provide flight-software-compatible support for functions on
              which the other packages rely
          * bp (bundle protocol), an implementation of the Delay-Tolerant
              Networking (DTN) architecture's Bundle Protocol.
          * dgr (datagram retransmission), a UDP reliability system that implements
              congestion control and is designed for relatively high performance.
          * ltp (licklider transmission protocol), a DTN convergence layer for reliable
              transmission over links characterized by long or highly variable delay.
          * ams - an implementation of the CCSDS Asynchronous Message Service.
          * cfdp - a class-1 (Unacknowledged) implementation of the CCSDS File
              Delivery Protocol.
          .
          This package contains the binary files.
      Homepage: https://ion.ocp.ohiou.edu/ [ohiou.edu]

  • Better figure out how to access "subspace" if you want to play Halo 27 with your friends on Titan. Or maybe quantum entanglement could accomplish the same thing?
  • ... to turn "First Post" into a whole new game.

  • I've no doubt they'll succeed with this, but if they are going to do remote robot control they are going to have to develop a very 'interesting' command structure. I can see it now... Go Forward... Stop... Oops!
    • > if they are going to do remote robot control they are going
      > to have to develop a very 'interesting' command structure. I
      > can see it now... Go Forward... Stop... Oops!

      NASA's robots are a bit more sophisiticated that that. It's closer to "Go over by the green rock" (not quite there yet, but close).

  • It sounds like a UUCP implemented at layer3.

  • Gaming must really suck, how are astronauts going to play online!? Well at least they can Download their Steam library..
  • Does anyone have more information about the Lego Mindstorms robot that was used in this experiment? I'd like to use it as an inspiration with the kids.

    The Curiosity Rover Made With Lego Mindstorms [wired.com] is pretty cool, but the fact that it uses "7 NXT Bricks, 13 NXT Motors, 2 Power Function Motors" makes it out of reach of the average home.

  • DTN is a store-n-forward protocol.

    Conceptually kinda sorta like email in that regard.

    The BP side of the equation brings the concept of bundling more information together in one unit (unlike IP, which tends to break info into smaller units , eg fragmentation).

    The plan being to bundle together all the information required for The Application to do the next thing.

    Imagine sending all the html-and-javascripts-and-css for a webpage in one (huge) packet. Your browser would have enough to render the page and s
  • This sounds just like FIDO used for inter-BBS and early internet email. Nodes would queue messages until the next hop was available. Why isn't the plan to use pairs on intangled particles to instantly pass information without regard for the speed of light or distance between end-points?
    • My layman's understanding is this:

      Because entanglement does not transmit state between the entangled pairs. It only allows you to, after separation, determine the state of the remote node by reading the local one.

      If you change the first node nothing is projected or expected to happen to the other in the pair.

      Believe me, if they find a way to transmit information FTL in any method it would be plastered all over the papers, the internet, and Slashdot, and would call into question many parts of science as it i

  • ...something amateur ("ham") radio operators have been using since the 1980's...

    -allen
    KC2KLC

  • Forget all this talk of UUCP, Fido and normal packet protocols, the closest current similarity is sending binaries over usenet.

    The most important part is the delay time, when you 'launch' a usenet message you won't receive anything at all from the remote end for a very long time. It will probably be long enough for you to transmit the entire message and then some.

    The medium also has some limitations ...

    • you can't send a 'message' over a few (hundred?) kilobytes, still small, but a lot larger than a si
    • The UUCP analogy is wrong anyway. The point is that even in UUCP days, we connected two modems, which transmitted data with X, Y or Z protocol. And that protocol was even more sensitive to propagation delays, because acks had to be sent much more frequently. Try set up UUCP to a mars probe, and you'll see that the layer 2 protocols will probably break down pretty badly. That's why DTN is really important and an entirely different kind of beast.
  • "Basically, a BP network — the one that will the Interplanetary Internet possible — moves data packets in bursts from node to node, so that it can check when the next node is available or up."

    Err... didn't this used to be called FidoNet [wikipedia.org]?

  • Legitimate Bufferbloat [wikipedia.org]?
  • Remember modem connections and 'feeds' for news and email? When most links were offline most of the time? Yep, networking before 'up all the time' connections were available to most of the world.

    I had a Linux (and before that a Mark Williams 'Coherent' UNIX like) computer that ran UUCP, and did dial on-demand connections. I had it download email and subscribed usenet 'news feeds' nightly from a local university that had 'free' connections for members of the local UNIX computer users group.

    It worked. It

...when fits of creativity run strong, more than one programmer or writer has been known to abandon the desktop for the more spacious floor. - Fred Brooks, Jr.

Working...