Re: [EToSat] Side-meeting on QUIC over high BDP ?

[Joerg Deutschmann <[e-mail]>]

Hi,

in our latest measurements [1] we black-box tested different QUIC implementations and different European satellite operators. The QUIC implementations are quicly and ngtcp (both draft-22) and Chromium QUIC (Q046). The operators are Astra, Avanti and Tooway. We also added one testbed setup based on dummynet.

We requested a single 1 Mbyte object and plotted time-sequence diagrams (a) without packet losses and (b) with 1% random packet loss. We decrypted pcap traces (quicly and ngtcp) or used log files (Chromium Q046) to plot offset numbers and see retransmissions. Packet losses have a very negative impact which was already mentioned at previous meetings. There are performance differences among implementations as well as operators. Please refer to [1] for the graphs.

I think testing different implementations and operators is fine, but these black-box tests are somehow unsatisfying because they do not explain *why* there are differences. Especially while the implementations are work in progress. A more detailed analysis would be great, but we currently have very limited time available for the QUIC over satellite topic.

[1] www7content.cs.fau.de/~deutschmann/KaConf2019/KaConf2019_QUICoverSatellite_paper.pdf
(section 4 and 5)

Best regards,
Joerg



On 09.10.19 17:14, Gorry Fairhurst wrote:
> Hi,
> 
> Could you share the topic with us (what version of QUIC; what aspect 
> were you testing) - so we can plan ahead and coordinate.
> 
> Gorry
> 
> On 09/10/2019, 16:07, Joerg Deutschmann wrote:
>> Dear Nicolas,
>>
>> I'd like to join but probably can't make it to Singapore. We have some 
>> limited results which I just sent to this mailing list...
>>
>> Best regards,
>> Joerg
>>
>>
>>
>> On 03.10.19 10:10, Kuhn Nicolas wrote:
>>> Hi,
>>>
>>> We are planning a side-meeting at next IETF on “Performance evaluation
>>> of QUIC in various environments”.
>>>
>>> The main idea is to gather people evaluating QUIC when it comes to
>>> exploit paths with various characteristics – with a slight focus for
>>> high BDP networks.
>>>
>>> Evaluations of QUIC over satellite has been presented in multiple IETF
>>> sessions [1-5], but a dedicated session would enable to gather whoever
>>> is interested in this activity.
>>>
>>> Is anyone interested in presenting experimental results and/or attend
>>> such meeting ?
>>>
>>> Cheers,
>>>
>>> Nico, on the behalf of John, Gorry and Emile
>>>
>>> [1]
>>> https://datatracker.ietf.org/meeting/103/materials/slides-103-maprg-quic-and-satcom-nicolas-kuhn-01 
>>>
>>>
>>>
>>> [2]
>>> https://datatracker.ietf.org/meeting/104/materials/slides-104-maprg-satellite-internet-performance-measurements-jorg-deutschmann-01 
>>>
>>>
>>>
>>> [3]
>>> https://datatracker.ietf.org/meeting/105/materials/slides-105-maprg-measuring-quic-dynamics-over-a-high-delay-path-01 
>>>
>>>
>>>
>>> [4]
>>> https://datatracker.ietf.org/meeting/105/materials/slides-105-panrg-quic-over-satellite-00 
>>>
>>>
>>> [5]
>>> https://datatracker.ietf.org/meeting/105/materials/slides-105-panrg-quic-over-in-sequence-paths-with-different-characteristics-00 
>>>
>>>
>>>
>> _______________________________________________
>> EToSat mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/etosat
>