January 2013
doc.: IEEE 802.11-13/0126r3
802.11mc
HEMM
Date: 2013-01
Authors:
Name
Company Address
Phone
email
Graham Smith
DSP Group
916 358 8725
[email protected]
1037 Suncast
Lane, Ste 112,
El Dorado Hills,
CA95762
Alex Ashley
Submission
Slide 1
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
Abstract
This document contains a discussion and proposals relating to
CID148
What is HEMM anyway? What does "HCCA, EDCA
mixed mode" mean? Which elements of HCCA and
EDCA are used in HEMM?
Submission
Slide 2
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
HEMM
• Defined as “HCCA, EDCA Mixed Mode”
• Appears only 6 times in 11e but a lot more in 802.111012
• 2 Major places in Specification:
– Table 8-5 TID
• 8-15 is used for HCCA
• 8-15 is also used for HEMM
• (So no way to differentiate)
– Access Policy in TS Info Field
• 11 = HEMM
• 01 = HCCA
• THIS IS THE ONLY DIFFERENTIATION IN THE SPEC
Submission
Slide 3
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
HEMM
• TS Operation 10.4.1
– Access Policy “to modify the HC’s scheduling behavior”.
– A TSPEC request may be set so that both HCCA and EDCA
mechanisms (HEMM) are used.”
Q. How does a TSPEC specify both HCCA and EDCA mechanisms?
ANS. By setting Access Policy to 11.
Submission
Slide 4
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
HEMM – Last Para 9.19.3.5.1
Edited by me for clarity
“HCF contention-based channel access shall not be used to transmit MSDUs belonging to an
established TS (with the HC’s acceptance of the associated TSPEC), unless:
•
the granted TSPEC indicates it is permitted to do so when the Access Policy subfield of the TS
Info field is equal to “HCCA, EDCA mixed mode” (HEMM),
• the polled STA utilized the full TXOP provided by the HC, and it has more MPDUs to send. “
Discussion on this:
If there is a HCCA time slot allocated then the MSDUs must use it – that seems reasonable
BUT it stipulates that an TS cannot transmit using EDCA at all if the TSPEC is for HCCA.
If TSPEC is for HEMM then TS can use EDCA for any ‘surplus’ packets but only if time slot is
totally used up.
So, if TSPEC is HEMM, then excess packets can be sent using EDCA
If TSPEC is HCCA, then excess packets can not be sent using EDCA.
This raises the question of EDCA Admission Control if the AC has ACM bit set.
Submission
Slide 5
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
HEMM – Last Para 9.19.3.5.1 cont.
“When this QSTA sends frames belonging to a TS using contention-based channel
access, it shall encode the TID field in the QoS Data frame with the TID
associated with the TS. “
OK, no problem here, irrespective of HEMMS
When the QAP grants a TSPEC with Access Policy subfield set to HEMM and if
the corresponding AC needs admission control, the QAP shall include the Medium
Time that specifies the granted time for EDCA access in the ADDTS response
frame.”
So, IF ACM BIT SET, the AP returns a “Medium Time” for HEMM.
Question:
• What is value of the Medium time? It is only for “excess traffic” so no real
way of knowing what is required. Simple is to return a Medium Time equal to
the values used in the TSPEC but his would be inefficient.
Submission
Slide 6
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
So what is HEMMS?
• HEMMS is identified in a TSPEC by setting the Access
Policy to 11. The TID is set same as HCCA.
• HEMMS is set in the TSPEC so that the STA can send
surplus packets outside the time slot.
• A HEMMS TSPEC response will includes Medium
Time, if AC has ACM set.
• Questions
• “Do we need HEMM in addition to HCCA, or should we
allow HCCA similar access to EDCA?”
• “Why would an HCCA STA ever not select HEMMS?
Hence, why do we need two schemes?
Submission
Slide 7
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
Keep HEMMS?
• So HEMMS seems to be a way to protect an AC with ACM if there
are surplus packets.
When does this problem arise and is it a big enough problem?
Problem arises if traffic exceeds the Time Slot allocation
How often? How much?
More likely to be a problem for Video (VBR) and not a
problem for voice or audio (CBR) unless the medium gets
blocked.
As the medium is always possible to be blocked, one could
argue that HCCA should never be used, and only HEMM
makes any sense. If so, why have two schemes?
I personally cannot see any need for both schemes as
HEMM has the back up that would be silly to not have.
Submission
Slide 8
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
If we keep HCCA and HEMMS
• Need to clarify how the Medium Time is calculated and
to make clear if required if no AC has ACM bit set.
– Annex N could cover this
– What should the value of the Medium Time be?
A proportion? The full amount? Even if one time slot is blocked,
as Medium Time is calculated over a 1 second period, still not easy
to work out what proportion it should be. Also it should be small
as only used if the TSPEC ‘got it wrong”. (Would this effect the
admission policy calculation – HIGHLY UNLIKELY)
– Should the STA be encouraged to modify its TSPEC if it finds
itself having significant “excess” packets? No need if HEMM.
SO why use HCCA ever, if HEMM provides this safe haven?
why have two schemes?
Is the only real reason to keep both is so that the term “HEMM” can be used
(If HCCA did not exist, the “H” cannot be used)
Submission
Slide 9
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
Just have HCCA
• Simply allow HCCA to spill over into EDCA in the case that there
are surplus packets.
• We have argued that there is no practical reason why any TSPEC
would restrict itself to HCCA only and the safe path is always to
set HEMM as the Access Policy.
Questions
• Do we still need a Medium Time in the ADDTS Response?
Note – only required if AP supports HCCA AND EDCA Admission
Control.
– NO - If ACM bit set, simply allow excess packets to use AC on the
assumption that this is a rare event. Remember TSPECs are
pessimistic. Also if ACM not set, no problem, not needed.
– YES - Have a Medium Time so that STA can test its Used Time.
What Medium Time? How much? Does it make any difference to
ACM access?
Submission
Slide 10
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
HEMM – would we miss it?
In my opinion, NO
• no one would miss it.
• No need for two schemes for HCCA
Mark H has indicated that a STA could cheat EDCA
Admission Control by sending a ‘small’ HCCA TSPEC
and then using the ACM AC.
Whether this is a real benefit as STA must still must use
time slot, has a CFP allocation to itself, so why
deliberately move over to a contention channel and get
less efficiency?
BUT, worthy of consideration so please see Proposal B
Submission
Slide 11
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
Further Discussion and Argument
•
•
•
•
HCCA STAs best interest to send all its traffic during the CFP – it is way more
efficient and safer compared to EDCA (latency, etc) – hence what is incentive
for the STA to cheat and simply use EDCA ACM access when it can get CFP?
However, even in the case of the HCCA STA using the time slot, the possibility
of interference is always present, so restricting the HCCA STA in any way
could be dangerous.
So, if both options, the STA would still be better to ‘play it safe’ and use
HEMM just in case. If the AP is not using EDCA Admission Control, then
there is no disadvantage at all to selecting HEMM.
If the AP is using EDCA Admission Control, i.e. VO and VI are set to ACM, as
well as HCCA Admission Control, then the disadvantage is only for the AP in
that it now needs to grant some bandwidth in the form of Medium Time (as it
stands at the moment). The STA now has to not only service the time slot
allocations but should, in theory, also keep account of its EDCA Used Time (I
very much doubt if in practice it would, but in theory and to meet the
Standard, it should).
Submission
Slide 12
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
Discussion and Argument
•
•
•
•
Proposal for HCCA only with no medium time is to keep it simple. How
often an HCCA QAP would use EDCA Admission Control is unknown,
and theoretically zero, the overall bandwidth allocation for the AP is
almost certainly covered by the HCCA TXOP allocation
Given the choice of HCCA or HEMM a STA would and should always
choose HEMM as it is “fail safe”, hence it does not make sense to have
two schemes.
The argument for keeping both HCCA and HEMM is that it is the status
quo and what if a TSPEC is send with Access Policy 11 (HEMM). One
solution is to simply use 01 and 11 for HCCA,
Proposal to keep both still needs agreement on the Medium Time
recommendation for HEMM, and I suspect that may be hard and possibly
subject to many comments whatever value is chosen. But a Medium Time
recommendation cannot be ignored as it certainly is a gap in the
Standard.
Submission
Slide 13
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
Proposal A
• Remove all references to HEMM in the Standard
– 01 and 11 Access Policy both refer to HCCA in order to cover
legacy and backward compatibility.
• Rewrite the last para of 9.19.3.5 as follows:
“HCF contention-based channel access shall not be used to transmit MSDUs belonging
to an established TS (with the HC’s acceptance of the associated TSPEC), unless the
polled STA:
·
Utilized the full TXOP provided by the HC, and
·
At the completion of the TXOP provided by the HC the STA has more MPDUs
to send.
If both these conditions have been met, this STA may transmit MPDUs belonging to an
established TS using contention-based channel access, irrespective of the value of the
ACM bit in the corresponding access category. When this STA sends frames belonging
to a TS using contention-based channel access, it shall encode the TID subfield in the
QoS Data frame with the TID associated with the TS. If the STA transmits MSDUs
belonging to an established TS using contention based channel access, the STA should
request modification of the associated TSPEC to an HCCA schedule that allows this TS
to be serviced entirely using TXOPs in a CFP.”
Submission
Slide 14
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
Proposal B
• Retain HEMM in the Standard
• Leave the last para of 9.19.3.5 as is except last sentence:
“When the AP grants a TSPEC with the Access Policy subfield equal to HEMM
and if the corresponding AC needs admission control, the AP shall include the
medium time that specifies the granted time for EDCA access in the ADDTS
Response frame (see Annex N 2.2).”
• Add text to Annex N 2.2 that provides some recommendations on
what the Medium Time field should be in this case.
“N.2.2.1 HEMM Medium Time
When the AP grants a TSPEC with the Access Policy subfield equal to HEMM
and if the corresponding AC needs admission control, it is recommended that the
medium time that is included in the ADDTS Response frame has a value that is
1/10th of the value that is derived using the procedure described in N 2.2”
Note: the 1/10th value is open to discussion but what should it be?
Submission
Slide 15
Graham Smith, DSP Group
January 2013
doc.: IEEE 802.11-13/0126r3
Poll
• Proposal A or Proposal B?
Submission
Slide 16
Graham Smith, DSP Group