Re: Fwd: New Version Notification for draft-matsushima-stateless-uplane-vepc-00.txt

Hi, Satoru,

Satoru Matsushima wrote:
> On Wed, Jul 17, 2013 at 12:10 AM, Bruno Mongazon-Cazavet 
> <[email protected]> wrote:
> 
> 
> 	Le 16/07/2013 16:27, Peter McCann a écrit :
> 
> 
> 		Satoru Matsushima wrote:
> 
> 
> 			Hi Peter,
> 
> 			On 2013/07/16, at 3:49, Peter McCann <[email protected]> wrote:
> 
> 
> 				I guess I don't understand the "PDN Prefix" and "subnet ID" fields
> 				that you have in Figure 4.  You state:
> 
> 				   Each PDN is assumed to have single or several prefixes (called
> PDN 				   prefix) used to generate UE's address. Followed by the PDN
> prefix in 				   Figure 4, there is 16-bit TEID assigned for a UE's
> session at SGW on 				   the control plane.  TEID is 16 bits identifier
> in GTP header to 				   distinguish each bearer.  The remaining bits are
> filled by subnet ID. 				   The prefix is allocated per UE and used for
> address assignment by 				   SLAAC or DHCPv6. 				Is Figure 4 supposed
> to be the Next Hop or the Destination?  I assumed 				it was Next Hop
> because it has the TEID encoded in it.  However, some 				of the above
> paragraph leads me to believe it is about the UE's user 				plane IP
> address.
> 
> 
> 
> 			The figure 4 illustrates UE's prefix so it shows the destination.
> 			Encoding TEID is just a seed to create UE's prefix in the
> stateless-pd 			manner.
> 
> 
> 		Doesn't the TEID change on every eNB handover?  I thought the UE's
> prefix 		should remain constant across handovers.
> 
> 
> 
> 
> If you use stateless-pd rule for UE's prefix generation, the embedded 
> TEID is SGW assigned one. So the delegated prefix is stable during 
> across eNB handover. Please see the section 3.3 in update version of 
> the draft. Anycast routing no longer requires hand-over between SGW in 
> virtualized EPC.
> 
> 
> 
> 	Yes TEID changes on handover and when UE exits Idle-Mode (it has 
> previously entered).
> 
> 
> Thanks, we haven't mention about in the case of Idle-Mode exit. will 
> update with it in a next version. It would be in control-plane 
> awareness section.

I think taken together, these facts would mean that the TEID space is
shared across all EPC-E devices (they all have the same anycast address)
and also that a TEID would need to remain allocated even to idle UEs.

Would you eventually run out of TEID space?

> 			We hasn't illustrate yet how to encode the TEID in next-hop. 			But
> as we mentioned in the document, we expect BGP remote next-hop as 			the
> way to encode it in the sub-TLV of the remote next-hop attribute.
> 
> 
> 		I would think that the Next-Hop would change on mobility events, but
> the 		Destination would remain the same (UE prefix).  Why is this not
> the case?
> 
> 
> 
> Yes, exactly.

Ok, so the Destination IPv6 address needs to remain the same during the lifetime
of a UE session, even across idle transitions and mobility events, right?

-Pete