On Wed, Aug 5, 2009 at 12:50 PM, Shane Amante<[email protected]> wrote:
> Sam,
>
> On Aug 5, 2009, at 09:01 MDT, Sam Hartman wrote:
>>>>>>>
>>>>>>> "Shane" == Shane Amante <[email protected]> writes:
>>
>>   Shane> Take a look at the following URL:
>>   Shane> http://www.sixxs.net/faq/connectivity/?faq=ipv6transit
>>   Shane> (Note, I can't vouch for the accuracy of the entire list,
>>   Shane> but it's about the best publicly available list I've seen).
>>
>>   Shane> That list includes some very large networks.
>>
>>
>> Thanks for the data!  Naturally that's only part of the answer to
>> Margaret's question.  We know which companies offer native IPV6
>> transit service.
>> We don't know:
>>
>> 1) How much V6 traffic they have 2) Whether they have V6
>> infrastructure that would scale to their current V6 traffic or say
>> their current V4 traffic 3) How much is done in hardware for their V6
>> infrastructure.  4) How much of their network supports V6/how
>> widespread their V6 works.  Any information you can provide on these
>> sorts of questions would be very useful.
>
> I think question #1 is a little impractical, given that data is highly
> sensitive/confidential to each network.  However, one likely can Google/Bing
> around and find public Exchange points with graphs that show current v6
> traffic (as compared to v4 traffic).  Of course, I would highlight that
> these are *public* exchange points, thus they will not show v6 traffic going
> through private peering connections as well as staying Intra-provider,
> (i.e.: customer-to-customer inside a single provider) -- see first sentence
> about sensitivity/confidentiality.  That said, there have been presentations
> at the last couple of IETF's (I forget if it was in 'behave' or 'v6ops'
> WG's, or both) surrounding Free Telecom's rollout of '6rd' that show their
> v6 traffic, which has some astounding growth and traffic rates.  Regardless,
> one should consider the following graphs a absolute "floor" of v6 traffic
> and there is definitely more, in absolute terms, v6 traffic than is visible
> in these graphs.
> Here's one example from AMS-IX showing v6 traffic:
> http://www.ams-ix.net/technical/stats/sflow/?type=ipv6
> And, here are graphs showing Total traffic load, (I believe this includes
> the v6 traffic in the graph above, but am not 100% certain on that point):
> http://www.ams-ix.net/technical/stats/
> According to these graphs, v6 traffic is just a fraction of v4 traffic, but
> it is definitely growing and, FWIW, it's a lot more than Inter-AS multicast
> traffic as seen by other graphs at the AMS-IX Web site.  :-/
>
> With respect to #2, SP's have been mandating that they only buy v6-capable
> HW for the last /several years/ as part of the normal growth/replacement
> cycle of network equipment.  So, yes, this equipment should scale to support
> v6 traffic at the same, (or nearly the same), rates as v4 traffic today.
>
> #3: All v6 forwarding is done in HW for larger, "PE" or "P" class devices,
> which are the subject of this thread.
>
> #4: Depends on where the carrier is at in their deployment of v6.  However,
> I would refer you back to question #2, which is even if it's not technically
> offered at a given location today, it is only a matter of time given that
> the HW is already deployed to do so (and, has been deployed for several
> years).
>
>
> To bring this back up a level, while it's /possible/ to encourage vendors to
> adopt the IPv6 flow-label as input-keys to their hash-calculations for
> LAG/ECMP, it takes [a lot] of time to see that materialize in the field.
>  Practically, you're probably looking at somewhere between, at best, a 3  -
> 5 year window, before it will actually appear in live, production networks,

s/networks/hardware-shipped-by-vendors/

You may see 2-3 year cycle on new asics for this feature to appear...
given 1-2 years for haggling/bugs/blah it's safe to say 3-5 yrs before
hardware is on the shelf to purchase.

Look at jason schiller's presos at IETF/RAWS/RRG that show 5-7 year
lifecycle of equipment in a network, so add some years before shipped
and deployed hardware is in general coverage in a network. Call it 10+
before it's ubiquitous...

> given that it has to be prioritized for development at the vendor, tested,
> released in software, then re-tested by the SP and, finally, deployed.

and hopefully deployed in the hardware, since that's where this
decision is made today (in v4 and v6 by 5-tuple). Software routers are
dead in any large SP for an serious workload. (P or PE devices)

>  That's not an "easy" process that happens in the blink-of-an-eye.  That's
> not to say that we (SP's) should not "encourage" vendors to do this, anyway,
> (we are/will) however if LISP (or other protocols like it that depend on
> tunneling) quickly gain traction, we need a way to deal with these traffic
> flows in our networks today, without telling customers: "turn off protocol
> <FOO>, because we can't deliver your packets".  Perhaps one way to satisfy
> the parties in this conversation would be something along the following
> lines:
> - LISP, and other protocols that wish to use tunneling, adopt UDP-lite (or
> UDP w/ 0 checksums) as a MUST for near-term deployment;
> - LISP, and other protocols that wish to use tunneling, adopt IP-in-IPv6
> tunneling with a flow-label required in the outer IPv6 header as a "SHOULD"
> for medium- to long-term deployments.
> ... assuming vendors successfully adopt the IPv6 flow-label as input-keys in
> their hash calculations at some point in the future, we come back and
> deprecate the UDP/IPv6 tunnel method and elevate IP-in-IPv6 w/ flow-labels
> to a MUST.
>
> It likely wouldn't hurt to go back and "tighten up" RFC 3697, as well, in
> order that it offers more prescriptive behavior in several places, not least
> of which would be:
> 1)  How to deal with regular IP-in-IPv6 encapsulation, (i.e.: likely by
> creating a flow-label based on a "hash" of the incoming, to-be-encapsulated
> L3 & L4 header fields), since the RFC only discusses IPsec tunneling; and,
> 2)  Removing other "gems" (or clarifying them) like the second sentence in
> the following:
> ---cut here---
> IPv6 nodes MUST NOT assume any mathematical or other properties of the Flow
> Label
> values assigned by source nodes.  Router performance SHOULD NOT be dependent
> on the
> distribution of the Flow Label values. Especially, the Flow Label bits alone
> make
> poor material for a hash key.
> ---cut here---

'flow label bits alone make a poor material for a hash key'... isn't
this the reverse of saying that we'll (operators) require vendors to
use flow-label for hashing on ECMP/LAG? If so, then... I don't think
flow-label's going to cut it.

-chris

> ... specifically, if "router performance" is meant to imply the behavior of
> load-balancing on ECMP/LAG paths, then router performance *is* heavily
> dependent on the distribution of flow-label values ...
>
> I've said all I'm going to say on this thread ...
>
> -shane
> _______________________________________________
> lisp mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/lisp
>
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to