Re: [STDS-802-16-MOBILE] [Re] [STDS-802-16-MOBILE] [Handoff] Harmonization of 04_105_Enhac ed_HO_re-entry
Dear Chulsik and all
I agree on your comment 2).
Surely, the end-to-end TCP session is disconnected if the IP
re-establishment carried on during the TCP session,
or UDP packet is useless if the destination IP address is changed.
I have the contribution (IEEE C802.16e-04_76r2 Handover - Data forwarding
and deferring IP re-establishment.pdf)
regarding this problem.
In summary, in this contribution, I highlighted the fact that if there is
data for an MSS,
which is about to perform handover, to receive,
the IP re-establishment should be deferred and data is forwarded (tunneled)
from the serving BS to the target BS.
Please refer to my contribution for more details.
Thank you.
Best Regards,
Ronny (Yong-Ho) Kim
LG Electronics.
________________________________________
From: owner-stds-802-16-mobile@LISTSERV.IEEE.ORG
[mailto:owner-stds-802-16-mobile@LISTSERV.IEEE.ORG]
Sent: Thursday, June 17, 2004 8:31 PM
To: STDS-802-16-MOBILE@LISTSERV.IEEE.ORG
Subject: [STDS-802-16-MOBILE] [Re] [STDS-802-16-MOBILE] [Handoff]
Harmonization of 04_105_Enhac ed_HO_re-entry
Hello Itzik and all,
I feel some confusion about the recent HO Ad Hoc's discussions. If my
understanding is not right, please let me know that. I described my
understandings in the followings (These are not restricted on the 87r2
documents only.):
1) The invited ranging (or fast ranging) is not possible in principle
during HO in OFDMA PHY, except for the case of the pre-associated MSS
having the adjust parameters for the target BS.
If the cells between serving BS and target BS largely differs in the cell
coverage, then the timing adjustment must be needed to the MSS before it
communicates with the target BS. So in the OFDMA-PHY, the code-based
ranging should be carried on for adjusting the timing offset between the
cells. But in OFDMA-PHY, the message-based ranging cannot be certified to
appropriately adjustable the timing offset without UL preamble, and the
non-timing aligned messages in a certain subchannel will give the other
subchannels a large interference, therefore we will do the code-based
ranging for timing adjustment during initial network entry and HO.
I think the Code-based ranging should be preceded before the RNG-REQ
message transfer (for transmitting MSS MAC Address, etc) during the HO
process in OFDMA-PHY (which cannot support the message-based Ranging.)
If we can apply the invited ranging only in the case of pre-associated MSS,
then we should explicitly describe these in the HO texts.
2) During the HO process, the IP re-establishment shall not be
supported without disconnection of the end-to-end TCP sessions.
In my understanding, the end-to-end TCP session should be disconnected if
the IP re-establishment carried on during the TCP session. Because the five
parameters (the sender's IP address, the receiver's IP address, the
transport protocol number, the sending application's port number, and the
receiving application's port number) are assumed to identify a connection,
if any of these parameters change (for example, the node assigned a
different IP address), the connection will break (i.e., communication will
no longer be possible).
Thus, in the case of IP re-established, the connection cannot be maintained
and the handover cannot be supported. So, we should make the tunnel between
serving BS to target BS while the subnet changed by prohibiting the IP
re-establishment.
And in the Idle mode, we described it as a state of returning all the
resources such as CID and IP address, therefore so called Idle mode
handover is not a handover (because the connection is not maintained during
Idle handover), but only a location update state. In the Idle mode, the IP
re-establishment does not occurred because it did not have the IP address
from the first time in Idle mode.
So, we do not need to discuss the IP re-establishment during handover and
idle mode.
3) For support the Soft Handover and/or fast BS switching, in my
understanding, the current IEEE 802.16d specification should be changed
(because it is not supported by current PHY, especially for OFDMA-PHY). But
the current PAR of TGe have the scope of backward compatibility with IEEE
802.16d, so we cannot put in the SHO and fast BS switching idea to the TGe
specification before we changed the TGe PAR as acceptable for PHY change,
even in the case of backward-incompatible to the TGd specification for
fully support the mobility feature in the current TGe specification.
For the case of SHO, the carrier permutation mechanism for interference
averaging effect for different subchannels in different BSs cannot be
maintained. So we can use the Band AMC subchannel concept to the SHO case,
but then the BSs should know each neighbor BS's channel and CID allocation
status. It needs an supervised controlling entities for BSs such as
cellular mobile's BSC concept, and the current BSs should only be the
transceivers which have no MAC functions.
For the case of Fast BS switching, the dedicated uplink feedback channel
for the MSS in every BS is required to provide the rank and CINR value of
the neighbor BSs for every frame to the BSC or anchoring BS. It cannot be
provided by using the current message based mechanisms, because it has
inherent delays and high overhead for transmitting these information. In
cdma2000 1x EV-DO, it can be provided using the small overhead by using the
most preferred BS's Walsh covering and transmission of CQI value, but in
OFDM or OFDMA it cannot be easy to support.
To solve these problems, we can design the dedicated UL feedback channel,
but it will gives us a backward compatibility issue with IEEE 802.16d.
4) For CID field designing issue, the backward compatibility problem
may occur. In IEEE 802.16d specification, The BS-ID is constructed as a msb
24-bit of operator ID, and a lsb 8-bit of sector ID. So, the Jung-won's CID
concept which have the separated sector ID of 8 bits and FA-ID of 8 bits is
not compatible with TGd spec.
But, we can think about the TGd document's description more deeply. In TGd
specification, The sector ID concept is only described in the compressed
MAP of SCa and OFDMA-PHY, not in the DL-MAP which is the superset of
compressed MAP. In my understanding, the compressed DL-MAP described only
as a compression of original DL-MAP by reducing the unnecessary parameters
in DL-MAP. But, the compressed DL-MAP described the sector ID concept which
is not in original DL-MAP. And, also the sector ID concept is not described
anywhere in the TGd specification.
I think, on this issue, the TGd specification have the problem in nature,
so we have the right to consider of breaking the backward compatibility
with TGd specification.
If we do not change the concept of sector ID in TGd specification's
compressed DL-MAP, then we cannot put in the concept of current Handover Ad
Hoc's inter-sector handover in the same BS. Because the sector is naturally
a base station according to TGd. It will treat as a different type of
handover between BSs.
Sincerely,
Chulsik Yoon,
Senior Engineer, ETRI
원본 내용:
보낸사람: owner-stds-802-16-mobile@LISTSERV.IEEE.ORG[itzikk@RUNCOM.CO.IL]
받는사람: STDS-802-16-MOBILE@LISTSERV.IEEE.ORG
제목: [STDS-802-16-MOBILE] [Handoff] Harmonization of
04_105_Enhaced_HO_re-entry and 04/87r1
받은날짜: 2004/06/17 목 01:45
Resend with correct subject and handoff tag!
Itzik.
------------------
Hello all,
I have uploaded document named "C802.16e-04_87r2" to the handoff ad hoc
upload directory.
It is my attempt to harmonize documents "C802.16e-04_105_Enhanced
Handover Re-entry" and "C802.16e-04_87r1"
I have used the descriptive text of C802.16e-04_87r1 and added
informative and normative definitions of Communication shared levels
concepts from "C802.16e-04_105_Enhanced Handover Re-entry"
One point though, I don't agree with the concept which was presented at
C802.16e-04_105_Enhanced Handover Re-entry of changing the RNG-REQ
message and adding to it TLVs with different processes.
I think that such approach breaks communalities of MAC behavior with TGd
and other MAC state machines.
The approach that was taken is to clearly define the communication
sharing level, and which procedure is required according to the working
level.
The optimization of the process is done in two levels; first, all the
network entry messages may be transmitted in a consecutive way, in one
frame (assuming to have enough allocation from the BS).
Second, according to the sharing level, some of the network entries may
be skipped. I agree that this adds additional MAC behavior, but it does
it in a high level and not changes the local behavior per process (e.g.
state machine).
I did not receive any additional inputs from Phil and Jung-Won Kim, and
I'm using this medium to seek for comments from the entire ad hoc group.
Thanks,
Itzik.