[Sis-dtn] R: ION and in-order delivery

Sun Feb 18 18:18:19 UTC 2024

Just a bit of additional information. We can also have duplicated bundles not only out of order bundles, and this even in the absence of custodial reTx, due to redundant retx at CL.  Thus, an application must be robust against them.
In IP I suppose it holds the same. IP packets are never retransmitted by IP, but frames (containing an IP) could.
The application needs either to be "idempotent", or to be able to discard duplicate bundles autonomously, e.g. by checking Source EID and timestamp and comparing it with a list of already received bundles.
Yours
 Carlo

________________________________
Da: SIS-DTN <sis-dtn-bounces at mailman.ccsds.org> per conto di Keith Scott via SIS-DTN <sis-dtn at mailman.ccsds.org>
Inviato: domenica 18 febbraio 2024 18:42
A: Scott Burleigh <sburleig.sb at gmail.com>
Cc: sis-dtn at mailman.ccsds.org <sis-dtn at mailman.ccsds.org>; 구철회 <chkoo at kari.re.kr>
Oggetto: Re: [Sis-dtn] ION and in-order delivery

A very quick search yielded a couple studies that claim anything between 3 and 56 percent (yes, quite a range,  but I can't really review things right now: in a car).

Remember also that many things we think of as Streaming (eg Netflix, youtube) use tcp.

However. I think the points are:

1. IP can reorder packets
2. BP can reorder packets

As a consequence, applications should not assume that packets / bundles will arrive in transmission order.

Another consequence of 2 is that even if YOUR network maintains fifo order, somebody else's may not.

    Keith

On Fri, Feb 16, 2024, 10:05 PM <sburleig.sb at gmail.com<mailto:sburleig.sb at gmail.com>> wrote:

Right, that’s why I said “rare”.  That is, from what I’ve been hearing about streaming I’m guessing it’s relatively rare.

Scott

From: Keith Scott <keithlscott at gmail.com<mailto:keithlscott at gmail.com>>
Sent: Friday, February 16, 2024 9:08 AM
To: sburleig.sb at gmail.com<mailto:sburleig.sb at gmail.com>
Cc: 구철회 <chkoo at kari.re.kr<mailto:chkoo at kari.re.kr>>; sis-dtn at mailman.ccsds.org<mailto:sis-dtn at mailman.ccsds.org>
Subject: Re: [Sis-dtn] ION and in-order delivery

But IP will deliver things out of order, just for different reasons.  Generally it's due to routing or link bonding or something like that.

    --keith

On Fri, Feb 16, 2024 at 3:25 PM <sburleig.sb at gmail.com<mailto:sburleig.sb at gmail.com>> wrote:

My sense is that this issue would never have come up if the implementation of bplib had not optimized for storage management by storing bundles in non-contiguous locations (as space became available) but clocking them out in ascending location order rather than in reception/storage time order.

This is behavior that you’ll never see in IP because IP never stores packets at all, it simply forwards them immediately upon reception.  In this sense IP itself is non-delay-tolerant, i.e., it doesn’t compensate for unplanned lapses in connectivity by retaining data for transmission when connection is restored.  The mis-ordering of data will be routine in the bplib implementation of BP but (suitably) rare in any implementation of IP.

We call BP a store-and-forward or store-carry-forward protocol, but there actually isn’t any language that I can find in RFC 9171 that requires that bundles be stored at all; we just assume that we all know it has to happen.  That’s an assumption that a developer picking up the specification might not share, so some sort of note pointing this out might be helpful.

Scott

From: SIS-DTN <sis-dtn-bounces at mailman.ccsds.org<mailto:sis-dtn-bounces at mailman.ccsds.org>> On Behalf Of Keith Scott via SIS-DTN
Sent: Thursday, February 15, 2024 11:11 PM
To: 구철회 <chkoo at kari.re.kr<mailto:chkoo at kari.re.kr>>
Cc: sis-dtn at mailman.ccsds.org<mailto:sis-dtn at mailman.ccsds.org>
Subject: Re: [Sis-dtn] ION and in-order delivery

Carlo: yeah, it's a very simple scenario; I took an existing test and plotted bundle sequence number against receive order.

Cheol + Carlo: yes, I believe the misordering behavior is entirely allowed by the sentiment and letter of and specification.  The way I understand BP, the service provided by BP is similar to that provided by IP:

  - Delivery of (bundles / datagrams) to the destination (name / address)

In particular, in both BP and IP, (bundles / datagrams) may be lost, duplicated, delayed, and/or misordered during delivery.

Context

=======

Sorry I didn't provide context for the first message.  We were discussing in the SIS-DTN telecon implementations that might want to impose a requirement along the lines of:

    "All other things being equal, a bundle node should forward bundles in FIFO order."

That's not a requirement in any of the BP specs, but in general it's probably beneficial to applications if things don't arrive too much out of order if they don't have to.  Really this stems from folks who want to be able to:

    - In bpv6, set timers like custodial retransmission timers or timers associated with DTPC, etc.

    - Do something intelligent with application-layer timers related to lost chunks of a stream

I think the above guidance to implementers is probably good, though there's (IMHO) a world of complexity hiding under "All other things being equal", especially when we consider QoS and the group's desire to be able to dynamically modify QoS treatments of bundles via policy.

    --keith

On Fri, Feb 16, 2024 at 1:40 AM 구철회 <chkoo at kari.re.kr<mailto:chkoo at kari.re.kr>> wrote:

Hi, Keith.

I think that's the normal operational behavior in LTP and TCPCL even in the absence of packet loss during connection because there are discontinuity events between A-B and B-C.

When a discontinuity event occurs during a transaction of a LTP session, the LTP session could possibly not be ended successfully and may pause to await the Report segment from the receiving entity. Upon reconnecting, the affected LTP session must resume interrupted session processing (e.g., RS-DS retransmit-RS-RAS), while other LTP sessions can continue smoothly.

Considering that an LTP session typically holds a bundle, in this scenario, out-of-order bundle delivery can occur even without packet loss. *NOTE* Discontinuity effectively simulates packet loss. TCPCL just acts like LTPCL. I think you could check it in a wireshark captured packet by seeing a Report segment.

In your scenario, discontinuity events occurred 8 times (4 times for A-B, 4 times for B-C). I can observe 8 times of bundle arrival fluctuation events in the figure.

Maybe there could be other issues originating this phenomenon, but above is my thought. HTH.

Best,
Cheol

----------------------------------------------------------------------

Message: 1
Date: Thu, 15 Feb 2024 20:59:24 +0100
From: Keith Scott <keithlscott at gmail.com<mailto:keithlscott at gmail.com>>
To: sis-dtn at mailman.ccsds.org<mailto:sis-dtn at mailman.ccsds.org>
Subject: [Sis-dtn] ION and in-order delivery
Message-ID:
        <CAHdkBBmXz6sQrY50p6Ob3ThO005132Ph9KZOiSgW+=C-8J5Y0w at mail.gmail.com<mailto:C-8J5Y0w at mail.gmail.com>>
Content-Type: text/plain; charset="utf-8"

I ran a quick test with a 3-node ION network A-B-C where for the first minute A is connected to B (via LTP) and B is connected to C (via TCP).
After that the topology repeats every two minutes; during the first minute A is connected to B and during the second B is connected to C.  At t=540s full connectivity (A-B and B-C) is restored for one minute.  There is no artificially induced loss in this scenario.  The max aggregation size for LTP was 100kbytes and the max aggregation time was 1s, e.g.:
*    a span 2 10 10 64000 100000 1 'udplso 10.44.3.2:1113<http://10.44.3.2:1113>
<https://protect2.fireeye.com/v1/url?k=ce37f0ae-91ac9aa6-ce328120-ac1f6bdccbcc-797d3695053eff4c&q=1&e=3ae2454f-c7be-47fb-9265-5a883d8026a6&u=http%3A%2F%2F10.44.3.2%3A1113%2F> 1000000'*

[image: image.png]

I ran bping (will do a unidirectional test later).  The chart below shows the received bping sequence number as a function of the order in which bpings were received.

[image: image.png]

So yeah, it looks like ION will occasionally misorder bundles.  I think that's not ideal, but I strongly believe that it is compliant with both the spec and the intent of the service (bundle delivery) BP purports to provide.  (And most bundles were in order, which is a feat considering that between 60 and 540s the network is never end-to-end connected).

    --keith
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://protect2.fireeye.com/v1/url?k=f133e8de-aea882d6-f1369950-ac1f6bdccbcc-aa39b8d587d0d7dd&q=1&e=3ae2454f-c7be-47fb-9265-5a883d8026a6&u=http%3A%2F%2Fmailman.ccsds.org%2Fpipermail%2Fsis-dtn%2Fattachments%2F20240215%2F70f735de%2Fattachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 48080 bytes
Desc: not available
URL: <https://protect2.fireeye.com/v1/url?k=5a238e22-05b8e42a-5a26ffac-ac1f6bdccbcc-ee06a4af1da35812&q=1&e=3ae2454f-c7be-47fb-9265-5a883d8026a6&u=http%3A%2F%2Fmailman.ccsds.org%2Fpipermail%2Fsis-dtn%2Fattachments%2F20240215%2F70f735de%2Fattachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 79905 bytes
Desc: not available
URL: <https://protect2.fireeye.com/v1/url?k=c83cc342-97a7a94a-c839b2cc-ac1f6bdccbcc-17dd8d887bfde0f4&q=1&e=3ae2454f-c7be-47fb-9265-5a883d8026a6&u=http%3A%2F%2Fmailman.ccsds.org%2Fpipermail%2Fsis-dtn%2Fattachments%2F20240215%2F70f735de%2Fattachment-0001.png>

------------------------------

Subject: Digest Footer

_______________________________________________
SIS-DTN mailing list
SIS-DTN at mailman.ccsds.org<mailto:SIS-DTN at mailman.ccsds.org>
https://protect2.fireeye.com/v1/url?k=ac25f3ab-f3be99a3-ac208225-ac1f6bdccbcc-2472f80e5d5dbacd&q=1&e=3ae2454f-c7be-47fb-9265-5a883d8026a6&u=https%3A%2F%2Fmailman.ccsds.org%2Fcgi-bin%2Fmailman%2Flistinfo%2Fsis-dtn

------------------------------

End of SIS-DTN Digest, Vol 152, Issue 3
***************************************
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.ccsds.org/pipermail/sis-dtn/attachments/20240218/c0a44bfd/attachment-0001.htm>