[LAC-TF] measuring ipv4 x ipv6 latencies

[Alejandro Acosta]

Hi All,
  One more  issue that I've heard which increments IPv6 delays is the
not hardware processing. AFAIK more routers/switch implements ASICs
for IPv4 but not for IPv6, therefore all routing & switching must go
to the OS and be processed on it.

Alejandro,

2011/4/20 Arturo Servin <aservin at lacnic.net>:
>
>
>        I presume that some of the problems that we will see for IPv6 compared with v4 it would be some paths with tunnels but I bet more for a difference in peering relationships. For example networks with low delay in v4 because a local peering and high delay in v6 for the lack of it. Fortunately the raw data is there, so anybody could start doing the analysis even without deploying probes.
>
>        Also, commenting about Alejandro's reply, I think that the main problem that he sees is 6to4 and Teredo (see Geoff analysis:  ). That should not happen with the probes that are being deployed for this project (but it might be good to take a look at the worst cases).
>
>
> Regards,
> .as
>
> On 20 Apr 2011, at 12:55, Antonio M. Moreiras wrote:
>
>> Hi Alejandro.
>>
>> I will try to follow your advice. In our case, the measurement points
>> are inside big networks (ISPs, universities, etc), so I think there will
>> be no cases of using Teredo or 6to4. But some of them can be running BGP
>> over tunnels with HE. I will try to consider it in the analysis.
>
>>
>> When I have some preliminary analysis done, I will let you know. If not,
>> we can compare the data and analysis in LACNIC meeting.
>>
>> Wouldn't it be possible also collect some data for us from BT network?
>>
>> --
>> Antonio M. Moreiras
>> Supervisor de Projetos
>> Núcleo de Informação e Coordenação do Ponto BR - NIC.br
>> moreiras at nic.br
>> (11) 5509-3553 - (11) 9466-2277
>> inoc-dba: 22548*amm
>> http://nic.br - http://ceptro.br
>>
>>
>> Em 20-04-2011 01:35, Alejandro Acosta escreveu:
>>> Hi Antonio,
>>>  Nice to hear about your study.
>>>  I wanted to tell you that I've running a similar research and I've
>>> found the same behavior, IPv6 latencies are around 35% worst than
>>> IPv4.
>>>  I got an advice for you, please if possible try to split the results
>>> between IPv6 native connections and IPv6 using transition methods
>>> (like 6to4 & Teredo), I also do have the data but I haven't done the
>>> maths! (sorry). I will be showing my results on Lacnic XV.
>>>
>>> Thanks in advance.
>>>
>>> Alejandro,
>>>
>>> 2011/4/19 Antonio M. Moreiras <moreiras at nic.br>:
>>>> Hello.
>>>>
>>>> We are taking some measurements of IPv4 and IPv6 latencies in the
>>>> Internet, against some Web sites and NTP sites.
>>>>
>>>> With this, we are trying to understand the current situation. In some
>>>> isolated tests, it seems that the IPv6 latencies could be 30% to 60%
>>>> worst than the IPv4 latencies, from some brazilian networks, including
>>>> ours. However, we are not sure about it.
>>>>
>>>> This test will integrate with (LACNIC's) SIMON in some near point in the
>>>> future. And probably we will discuss some results at IXPs meeting in the
>>>> next LACNIC.
>>>>
>>>> We are taking these measurements from our network, at NIC.br, and from
>>>> some other brazilian sites, but we would like to have other partners in
>>>> this test, from worldwide, and specially from LAC region, what would
>>>> facilitate the comparisons.
>>>>
>>>> You could see some preliminary data in this website:
>>>>
>>>> http://labs.ceptro.br/simonv6 (work in progress).
>>>>
>>>> Maybe you could host a measurement point for this?
>>>>
>>>> The measurements are taken by a java software, that runs also an
>>>> external traceroute. It runs preferencially in a linux/ubuntu 10.10 or
>>>> 10.04.
>>>>
>>>> Some of the participants are installing the software themselves. Others
>>>> are providing us with virtual machines. Both models are welcome.
>>>>
>>>> Organizations that agree to take part in this test, should also agree in
>>>> allowing us to share the results publically. The data to be publically
>>>> released is:
>>>> - orig ipv4, dest ipv4, delay ipv4, traceroute ipv4 (IP list)
>>>> - orig ipv6, dest ipv6, delay ipv6, traceroute ipv6 (IP list)
>>>>
>>>> Other information, as the AS number, location, upstreams, etc, can be
>>>> infered from these data.
>>>>
>>>> If you can help us with this, please let me know in pvt and I will send
>>>> further information.
>>>>
>>>> Regards!
>>>> --
>>>> Antonio M. Moreiras
>>>> IPv6.br project coordinator
>>>> Brazilian Network Information Center - NIC.br
>>>> moreiras at nic.br
>>>> +55 11 5509-3553 -  +55 11 9466-2277
>>>> inoc-dba: 22548*amm
>>>> http://nic.br - http://ceptro.br
>>>>
>> _______________________________________________
>> LACTF mailing list
>> LACTF at lacnic.net
>> https://mail.lacnic.net/mailman/listinfo/lactf
>
> _______________________________________________
> LACTF mailing list
> LACTF at lacnic.net
> https://mail.lacnic.net/mailman/listinfo/lactf
>