Re: [Detnet-dp-dt] [EXT] DETNET Synchronization Slides

Hi Jouni, 

That is a fair point. If only the PW nodes function as TCs, then indeed the accuracy will be lower.
I added a clarification about this in slide 6.

The updated slides are attached.

Thanks,
Tal.

>-----Original Message-----
>From: Jouni Korhonen [mailto:jouni.korhonen@broadcom.com]
>Sent: Thursday, February 09, 2017 10:35 PM
>To: Tal Mizrahi
>Cc: detnet-dp-dt@ietf.org
>Subject: Re: [EXT] [Detnet-dp-dt] DETNET Synchronization Slides
>
>Tal,
>
>I understood that in 2) the TC function would only be in the PW nodes we define
>leaving all intermediate “normal” LSRs between PW nodes as a source of higher
>inaccuracy. If this was not a correct understanding then the value in 2) is likely
>too high.
>
>- Jouni
>
>
>> On 08 Feb 2017, at 23:22, Tal Mizrahi <talmi@marvell.com> wrote:
>>
>> Hi Jouni,
>>
>> Sorry for not explaining the accuracy numbers on slide 6 further. I made some
>changes following the discussion, and also added notes saying that the accuracy
>will depend on the network topology and on the number of hops.
>>
>> Regarding approach 2, I believe sub-microsecond accuracy can certainly be
>achieved using transparent clocks in a real network. There are quite a few
>publications that show this. Here are a couple:
>> - For example, China Mobile have revealed in a keynote lecture in ISPCS 2014
>that they have measured an accuracy of 266 ns over 30 hops of transparent
>clocks. Unfortunately, this presentation is not in the public domain, but a short
>version of it is available  (https://www.ietf.org/mail-
>archive/web/tictoc/current/pdfp9ceHphbcE.pdf).
>> - Another example:
>http://wwwen.zte.com.cn/endata/magazine/ztecommunications/2010Year/no3
>/articles/201009/t20100913_191838.html
>> But again, this entirely depends on the nature of the network.
>>
>> Regarding approach 1, the accuracy will entirely depend on the network
>topology (the number of hops, the physical distance, the type of intermediate
>nodes, the delay variation). Hundreds of microseconds is just an example. Some
>of the research I took part in showed that in a multi-site public cloud network you
>can achieve a precision on the order of hundreds of microseconds using PTP
>without on-path support
>(https://sites.google.com/site/talmizrahihome/files/TaaSInfocomFinal.pdf?attred
>irects=0). But again, this entirely depends on the nature of the network.
>>
>> Please let me know if there are further comments.
>>
>> Thanks,
>> Tal.
>>
>>
>>> -----Original Message-----
>>> From: Jouni Korhonen [mailto:jouni.korhonen@broadcom.com]
>>> Sent: Wednesday, February 08, 2017 8:15 PM
>>> To: Tal Mizrahi
>>> Cc: detnet-dp-dt@ietf.org
>>> Subject: [EXT] Re: [Detnet-dp-dt] DETNET Synchronization Slides
>>>
>>> External Email
>>>
>>> ----------------------------------------------------------------------
>>> Thanks Tal,
>>>
>>> Didn’t we also agree adjusting the slide 6 “typical accuracy” numbers:
>>> 1) <1ms or hundreds of us
>>> 2) <tens of us
>>> 3) <1us
>>>
>>> - Jouni
>>>
>>>
>>> --
>>> Jouni Korhonen, Broadcom Ltd.
>>> M: +1-408-391-7160
>>>
>>>> On Feb 7, 2017, at 10:51 PM, Tal Mizrahi <talmi@marvell.com> wrote:
>>>>
>>>> Hi,
>>>>
>>>> The slides from today’s weekly meeting are attached.
>>>> I slightly updated them after the meeting based on the comments.
>>>>
>>>> Cheers,
>>>> Tal.
>>>> <Synchronization in
>>> DETNET.pptx>_______________________________________________
>>>> Detnet-dp-dt mailing list
>>>> Detnet-dp-dt@ietf.org
>>>> https://www.ietf.org/mailman/listinfo/detnet-dp-dt
>>

Re: [Detnet-dp-dt] [EXT] DETNET Synchronization Slides

Attachment: Synchronization in DETNET.pptx