[Sis-dtn] Positive reception claim vs. Negative reception claim in LTP Report Segment preparation and processing
구철회
chkoo at kari.re.kr
Tue Apr 5 06:06:17 UTC 2022
Let me guess.
I think populating the ACK/NACK reception claim information at the first time in which segmentation is lost is not quite different between them. Some differences can be observed during processing retransmitted data segments. I suppose there will be some differences in internal processing on a hardware (or similarly for software).
Consider below structure differences each ACK and NACK. When it encounters discontinuity of segment streams, i.e., loss of segmentation, by N times, (1) requires N+1 information set to hold the positive claims, and (2) requires N information set.
<<(1) normal ACK>> <<(2) NACK>>
+-- upper bound --+ +-- upper bound --+
(7000) (7000)
+---------------------+
positive | offset (3000) |
claim | length (4000) |
+----------------------+
+---------------------+
-- missing segments - negative | offset (1000) |
claim | length (2000) |
+----------------------+
+---------------------+
positive | offset (0) |
claim | length (1000) |
+----------------------+
1. (0)
+-- lower bound --+ +-- lower bound --+
We can just get exactly the same information of positive claims and negative claims from these two different structures. Consider a retransmitted segmentation (e.g., {1000, 1000}) is arrived,
(1)’s logic has to load two information set into memory for comparison, e.g., {0, 1000}, {3000, 7000}, and then start comparing two value of (0+1000) and (3000) to check if the newly accepted segment is located in the missing segment area. So, (1)’s logic will require 1) memory access cycles to bring two information set into registers, 2) ALU cycles to sum-up 0+1000, 3) and comparison cycles to compare it {1000, 3000} with the segment size of 1000~2000.
(2)’s logic only has to load one information set for the comparison, i.e., {1000, 2000}. So, (2)’s logic will require 1) memory access cycles to bring one information set into registers, 2) ALU cycles to sum-up 1000+2000, 3) comparison cycles to compare it {1000, 3000} with the segment size of 1000~2000.
If there are N positive/negative claims stored in memory, the process will iterate until it finds a matched area. So in localizing the segment area to be updated (2) requires shorter cycles than (1).
If a missing segment are fully filled by retransmitted segments, the current positive claims and negative claims have to be adjusted; (1) requires cycles for deleting one information set and updating another one and (2) requires cycles for just deleting current information set only. On the contrary, if the received retransmitted segment is located just mid of the missing segments area so that one additional information set has to be created, e.g., {1500, 500} is received, (1)’s logic requires cycles only for creating two register block and adjusting linked-list structure, but (2)’s logic requires additional cycles for updating old information set over cycles for creating one additional information set, but it happens just one time.
For a conclusion, I think (2) has benefits when the reception claims are very fragmented. Of course that’s just quick thoughts and more detail analysis is necessary for future.
Cheol
From: Felix.Flentge at esa.int <Felix.Flentge at esa.int>
Sent: Monday, April 4, 2022 8:22 PM
To: Tomaso.deCola at dlr.de
Cc: 구철회 <chkoo at kari.re.kr>; Jeremy.Mayer at dlr.de; sis-dtn at mailman.ccsds.org
Subject: RE: [Sis-dtn] Positive reception claim vs. Negative reception claim in LTP Report Segment preparation and processing
Hi,
yes, I also assume that for typical space links they would be quite similar in terms of efficiency. I think the question is also about implementation complexity: is it 'easier' to implement NAK-based re-transmission at high-data rates in hardware with maybe limited resources?
Regards,
Felix
From: <Tomaso.deCola at dlr.de<mailto:Tomaso.deCola at dlr.de>>
To: <Jeremy.Mayer at dlr.de<mailto:Jeremy.Mayer at dlr.de>>, <Felix.Flentge at esa.int<mailto:Felix.Flentge at esa.int>>, <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>>
Date: 04/04/2022 12:35
Subject: RE: [Sis-dtn] Positive reception claim vs. Negative reception claim in LTP Report Segment preparation and processing
________________________________
Hi All,
I tend to agree with Jeremy, from a pure ARQ effectiveness view point, using ACK or NACK for signalling (or detecting) losses pretty much depends on the channel model you are assuming underneath. I’m quite sure in the scientific literature you can find many papers about using either approach. For typical space links, probably using ACK or NAK does not bring significant differences from a performance standpoint. As to the optical and Ka-band communication links, again, I’d say it depends on the channel model and more concretely on how packet losses are distributed after channel coding and CRC control at frame level. In particular, it may depend on the specific reliability measures implemented at the physical layer (e.g., long interleavers, long codewords, etc…), hence possibly resulting in an almost error-free channel (with some sporadic erasures) or in a more correlated loss pattern. At the end, I don’t think we can come up with an ideal ARQ solution that works at best for all possible channels…
My 0.02 cents,
Tomaso
From: SIS-DTN <sis-dtn-bounces at mailman.ccsds.org<mailto:sis-dtn-bounces at mailman.ccsds.org>> On Behalf Of Jeremy Pierce-Mayer via SIS-DTN
Sent: Montag, 4. April 2022 12:09
To: Felix.Flentge at esa.int<mailto:Felix.Flentge at esa.int>; 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] Positive reception claim vs. Negative reception claim in LTP Report Segment preparation and processing
Hi Cheol, Felix,
The efficiency of positive vs. negative claims is highly dependent upon the behaviour of the underlying link. If a link has long periods of successful communication punctuated by brief (complete) fading events, then NACK may be better. If a link is more erratic, then the calculations become a bit harder and are highly dependent on the ratio and duration of successful vs lost packets/frames.
In most “reliable” space links, fading is pretty intermittent (until your elevation reduces), so ACK/NACK should be pretty similar. I think Ka/optical might upset this balance though… We’ll see.
Thanks,
Jeremy
From: SIS-DTN <sis-dtn-bounces at mailman.ccsds.org<mailto:sis-dtn-bounces at mailman.ccsds.org>> On Behalf Of Felix Flentge via SIS-DTN
Sent: Monday, April 4, 2022 12:08 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] Positive reception claim vs. Negative reception claim in LTP Report Segment preparation and processing
Ah, yes, of course you are right.
We will look into the negative ACK as part of our LTPv2 prototyping activity.
Regards,
Felix
From: "구철회" <chkoo at kari.re.kr<mailto:chkoo at kari.re.kr>>
To: <Felix.Flentge at esa.int<mailto:Felix.Flentge at esa.int>>
Cc: "sis-dtn at mailman.ccsds.org<mailto:sis-dtn at mailman.ccsds.org>" <sis-dtn at mailman.ccsds.org<mailto:sis-dtn at mailman.ccsds.org>>
Date: 04/04/2022 11:58
Subject: RE: Re: [Sis-dtn] Positive reception claim vs. Negative reception claim in LTP Report Segment preparation and processing
Sent by: chkoo at kari.re.kr<mailto:chkoo at kari.re.kr>
________________________________
Hi Felix,
I think current LTP spec quite works well with negative claim also. Consider below reception claim according to the LTP spec but negative claim.
lower bound = 0
upper bound = 7000
negative reception claim count = 1
offset = 1000
length = 2000
it means a receiver is requesting block of segements which starts at 1000 and length is 2000, i.e., 1000 ~ 2999, for retransmission.
A sender can safely remove 2 blocks, i.e., 0 - 999 and 3000 - 7000. I think it is simpler, lower overhead and *importantly* easier to calculate (acutally no painful for localizing the target segment position).
Cheol
--------- 원본 메일 ---------
보낸사람 : <Felix.Flentge at esa.int<mailto:Felix.Flentge at esa.int>>
받는사람 : "구철회" <chkoo at kari.re.kr<mailto:chkoo at kari.re.kr>>
참조 : "sis-dtn at mailman.ccsds.org<mailto:sis-dtn at mailman.ccsds.org>" <sis-dtn at mailman.ccsds.org<mailto:sis-dtn at mailman.ccsds.org>>
받은날짜 : 2022-04-04 (월) 17:40:24
제목 : Re: [Sis-dtn] Positive reception claim vs. Negative reception claim in LTP Report Segment preparation and processing
Hi Cheol,
interesting question. One thing I can think of is that the positive claims would allow you to free memory earlier while for negative claims you need to wait until the end of a session.
Regards,
Felix
From: "구철회 via SIS-DTN" <sis-dtn at mailman.ccsds.org<mailto:sis-dtn at mailman.ccsds.org>>
To: "sis-dtn at mailman.ccsds.org<mailto:sis-dtn at mailman.ccsds.org>" <sis-dtn at mailman.ccsds.org<mailto:sis-dtn at mailman.ccsds.org>>
Date: 04/04/2022 10:15
Subject: [Sis-dtn] Positive reception claim vs. Negative reception claim in LTP Report Segment preparation and processing
Sent by: "SIS-DTN" <sis-dtn-bounces at mailman.ccsds.org<mailto:sis-dtn-bounces at mailman.ccsds.org>>
________________________________
Greetings,
This is Cheol. I am developing an LTP reference implementation. During reading the LTP specification (RFC-5326), the preparation of reception claim in Report Segment makes me confusing about why it is positive claim not negative claim for segments that were not received successfully (i.e., NAK).
For reference, CFDP’s NAK PDU has the negative claim structure when it is requested to report missing PDUs. Does anyone know about the background of choosing the positive claim for NAK operation in LTP?
I think negative claim is simpler and more efficient in terms of overhead for sender and receiver both.
I like to listen experts’ opinion on LTP operation and honestly hope it to be changed in newly coming LTP spec.
Cheol
_______________________________________________
SIS-DTN mailing list
SIS-DTN at mailman.ccsds.org<mailto:SIS-DTN at mailman.ccsds.org>
https://mailman.ccsds.org/cgi-bin/mailman/listinfo/sis-dtn<https://protect2.fireeye.com/v1/url?k=933edf14-cca5b51c-933bae9a-ac1f6bdccbcc-93bc8ad36316533d&q=1&e=24a03daf-8e73-4317-a689-3216c529ea83&u=https%3A%2F%2Fmailman.ccsds.org%2Fcgi-bin%2Fmailman%2Flistinfo%2Fsis-dtn>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.ccsds.org/pipermail/sis-dtn/attachments/20220405/ca3121ce/attachment-0001.htm>
More information about the SIS-DTN
mailing list