Qos related issues in MAC and Baseline document #360

Slides:



Advertisements
Similar presentations
Doc.: IEEE /272a Submission June 2001 S. Choi, Philips Research Slide 1 Problems with IEEE (e) NAV Operation and ONAV Proposal Javier del.
Advertisements

IEEE CSMA/CA DCF CSE 6590 Fall /7/20141.
Doc.: IEEE /080r1 Submission January 2001 Jie Liang, Texas InstrumentsSlide 1 Jie Liang Texas Instruments Incorporated TI Blvd. Dallas,
Achieving Quality of Service in Wireless Networks A simulation comparison of MAC layer protocols. CS444N Presentation By: Priyank Garg Rushabh Doshi.
– Wireless PHY and MAC Stallings Types of Infrared FHSS (frequency hopping spread spectrum) DSSS (direct sequence.
1 QoS Schemes for IEEE Wireless LAN – An Evaluation by Anders Lindgren, Andreas Almquist and Olov Schelen Presented by Tony Sung, 10 th Feburary.
1 Dynamic Adaption of DCF and PCF mode of IEEE WLAN Abhishek Goliya Guided By: Prof. Sridhar Iyer Dr. Leena-Chandran Wadia MTech Dissertation.
IEEE Wireless LAN Standard. Medium Access Control-CSMA/CA IEEE defines two MAC sublayers Distributed coordination function (DCF) Point coordination.
Doc.: IEEE /097 Mechanisms for Transmission Suppression in January 18, 2001 Matthew Sherman, AT&T Labs - ResearchSlide 1 Mechanisms for.
Submission doc.: IEEE /569r1 November 2001 M. Benveniste -- AT&T Labs, ResearchSlide 1 An Access Mechanism for Periodic Contention-Free Sessions.
Doc.: IEEE /452 Submission December, 2000 Michael Fischer, Intersil Slide 1 A Hybrid Coordination Function for QoS Michael Fischer Intersil Corporation.
MAC Sublayer MAC layer tasks: – Control medium access – Roaming, authentication, power conservation Traffic services – DCF (Distributed Coordination.
MAC for WLAN Doug Young Suh Last update : Aug 1, 2009 WLAN DCF PCF.
Doc.: IEEE /361 Submission October 2000 Wim Diepstraten, LucentSlide 1 Distributed QoS resolution Greg Chesson-Altheros Wim Diepstraten- Lucent.
November 2000 Jin-Meng Ho, Texas InstrumentsSlide 1 doc.: IEEE /367 Submission p-DCF for Prioritized MAC Service Jin-Meng Ho, Sid Schrum, and.
Doc.: IEEE /248r0 Submission Bobby JoseSlide 1 February 2002 Contention Free TXOP Request and Allocation Issues Bobby Jose,
Multiple Access By, B. R. Chandavarkar, CSE Dept., NITK, Surathkal Ref: B. A. Forouzan, 5 th Edition.
Doc.: IEEE /110 Submission May 2000 Sunghyun Choi, Philips ResearchSlide 1 QoS Support in : Contention-Free MAC Perspective Sunghyun Choi.
Mathilde Benveniste Avaya Labs
Wireless MAC.
EA C451 (Internetworking Technologies)
Calibration using NDP Vincenzo Scarpa
Topics in Distributed Wireless Medium Access Control
Lecture 27 WLAN Part II Dr. Ghalib A. Shah
An Access Mechanism for Periodic Contention-Free Sessions
WUR coexistence with existing power save mode
How to collect STAs’ Tx demands for UL MU
Net301 lecture9 11/5/2015 Lect 9 NET301.
IEEE : Wireless LANs ALOHA, Slotted ALOHA
Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [Proposal for Collaborative BT and b MAC Mechanisms.
HCF medium access rules
Requirements and Implementations for Intra-flow/Intra-AC DiffServ
A Scheduling Scheme for Level-2 Enhanced PCF MAC Service
Fair and Protected DLS September 2007 Date:
Provision of Multimedia Services in based Networks
Multicast/Broadcast Communication With Acknowledge
Fair and Protected DLS September 2007 Date:
September 2010 Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [TG6 Proposed MAC comment resolution]
EDCA and BlockAck Extensions for Reliable Multicast/Broadcast Services
Enhanced MAC proposal for high throughput.
Protocol Details John Bellardo UCSD.
Hybrid Coordination Function (HCF) Frame Exchange and NAV Details
Calibration using NDP Date: Authors: December 2006
Burst Transmission and Acknowledgment
EDCF Issues and Suggestions
Terminology Corrections and Improvements for the TGe Draft
Comment resolution on BSR CID 8426
Peer Power Save Mode for TDLS
HCF medium access rules
Air Efficiency and Reliability Enhancements for Multicast
Comment resolution on CID 20175
Suggested changes to Tge D3.3
DL MU MIMO Error Handling and Simulation Results
HCF medium access rules
Acknowledgement for Multicast Streams
Proposed ERTS & ECTS Mechanisms
PCF Enhancements and Contention Free Bursts
Proposed ERTS & ECTS Mechanisms
Comment resolution on CID 20175
Suggested changes to Tge D3.3
HCF medium access rules
NAV Operation Rules under HCF
Scheduled Peer Power Save Mode for TDLS
Air Efficiency and Reliability Enhancements for Multicast
802.11g Contention Period – Solution for Co-existence with Legacy
QoS Metrics Date: Authors: January 2005 Month Year
HCF Channel Access And Inter-BSS Channel Sharing
Chapter 11 Comment Resolution for Letter Ballot 63
NAV Operation Rules under HCF
Indicating NGV Capabilities in MAC Header
Presentation transcript:

Qos related issues in 802.11 MAC and Baseline document #360 January 2001 Qos related issues in 802.11 MAC and Baseline document #360 Dr. Rajugopal Gubbi, Matthew Fischer and Dr. George Kondylis Broadcom, Corp.

What are the modifications? January 2001 What are the modifications? Providing distributed, but controlled channel access within CFP Resolving issues in 802.11-1999 to assist in Qos operation Resolving issues in Tx-ops in the current Qos-baseline document

Providing distributed, but controlled channel access within CFP January 2001 Providing distributed, but controlled channel access within CFP

January 2001 What is the problem? There were several proposals for EDCF and many papers Issue related to operating with legacy stations makes the proposals complex and not completely workable. None of the current EDCF proposals completely prevent collisions of frames of different priority Backoff after collisions needs to be different (compared to legacy algorithm) for the current EDCF proposals Hidden node and near-far effects worsens situation in all the current EDCF proposals Using +CF-Poll in CP does not solve this problem as the latency/jitter for those frames is not bound especially in the presence of legacy stations

Why not just use CFP for all traffic? January 2001 Why not just use CFP for all traffic? As defined today in 360r1, EAP/EPC can not always predict the traffic correctly when Multiple periodic streams are present, the CFP-rep-rate can not always be optimized for all traffic and all periodicity without wasting some bandwidth within CFP. And some QOS flows may be bursty or inefficient polling of ALL (E)STA low-latency, low-duty cycle, periodic traffic might also need faster beacon rate Hence there is a need to allow distributed, but controlled channel access for Qos data flow within parts of CFP

What does this paper discuss? January 2001 What does this paper discuss? This paper discusses the modifications to Qos baseline document. The modifications use the already accepted mechanisms, in 360r1, in an extended manner to achieve the goal of providing overall Qos with no increase in complexity to provide better efficiency.

What are the modifications? (1) January 2001 What are the modifications? (1) Use contention control within CFP with following generalization of its definition Generalize the definition of CC so that the ESTAs can send ANY pending control/management or Qos-data frame Use Duration/ID field of CC frame to indicate the end of controlled contention period (CCP) with b15=1 and b14=0 (as required in the current definition in Table-3). Provide a “CC-OP-limit” within CC frame as the max limit on the duration of frame transmission within CCP RR frames are solicited by EAP/EPC by limiting the CC-OP-limit to the duration of RR frame at the PHY rate used for CC Provide priority RANGE (px to py) to limit the class of traffic within CCP px=py means there is only one priority is allowed during CCP. Control and management frames can be sent in any CCP

Covers the entire CCP that follows the current CC frame January 2001 What are the modifications? (2) New format for CC Frame Max limit (in micro seconds) on the duration of data/control/management frame during the current CCP Covers the entire CCP that follows the current CC frame b15=1 and b14=0 as in Table-3 b6:b4 b2:b0 b3 b7 Lower limit on priority Upper limit on priority RSV Note: The Reserved octet can be used for CC-OP-limit field with values in multiples of 16 Hence save two octets in the CC frame

What are the modifications? (3) January 2001 What are the modifications? (3) Little change to the channel access method during CCP (almost same as currently defined in clause 9.2) All ESTAs use DIFS for channel access during CCP All ESTAs use the currently defined (clause 9.2.4) backoff algorithm during CCP. Change of transmission attempt from one priority frame to another is treated same as different MSDU attempt during CP (clause 9.2.4). Use of SLRC/SSRC and SRC/LRC. Retry procedure is same as that in 802.11-1999 (clause 9.2.5.3). The retry limits can be defined to be different for Qos-Data frames. CW counter for Qos-data is reset at the beginning every CCP to provide fair access. Otherwise, an ESTA that has backed off in an earlier CCP will be at disadvantage compared to the ESTA that constructed the frame in between two CCPs.

What are the modifications? (4) January 2001 What are the modifications? (4) CW is also reset when SLRC or SSRC reaches its limit, as defined in clause 9.2.4, to provide preference to ESTA that did not have transmission opportunity for a long time. NAV set in ESTAs, is applicable only for CP. During CCP ESTAs are allowed to transmit if priority range and CC-OP-limit are met CCP can end early if EAP/EPC grabs the channel after PIFS All ESTAs sense the channel becoming busy after PIFS and interprets that as the early-end of CCP, even though the received frame from EAP/EPC may be corrupted EAP/EPC always sends the CC frame with zero CCP duration when it ends the CCP early. AND uses PIFS for the following data+poll so that if the channel was free for a while the start of second frame from the EAP/EPC clearly signals the end of CCP RTS/CTS within CCP (as was proposed to be within CFP) to find out if the destination device is within range. NAV is not set by the ESTAs of the same QBSS. RTS is limited to cover the ONLY frame following the response CTS and ack, if any.

What are the modifications? (5) January 2001 What are the modifications? (5) Example of CFP operation CC with Zero CCP duration D1+X-Poll frame D1+X-Poll frame D1+X-Poll frame CC frame CC frame Early ending CCP CC frame Normally ending CCP CF-End Beacon TBTT TBTT Controlled Contention Period(CCP) PIFS Controlled Contention Period(CCP) SIFS PIFS PIFS PIFS PIFS SIFS Contention-Free Period(CFP) Contention Period(CP) EAP/EPC uses CF-Poll, Qos-CF-Poll, CF-Multi-Poll and CF-Sch to provide Tx-ops to ESTAs CCP: ESTAs use channel access mechanism described before for CFP

What are the modifications? (6) January 2001 What are the modifications? (6) Summary of the proposed changes All ESTAs recognize CC frame Limit transmissions as per CC-OP-limit indicated in CC-frame Limit transmissions to priority range indicated in CC frame Reset CW to CWmin at the beginning of CCP for the pending control/management/Qos-data frames Recognize the channel takeover by EAP/EPC after PIFS to signal the end of CCP EAP/EPC sends CC frame with zero duration to end the CCP early. And uses PIFS for the next data+poll frame. Use of RTS/CTS within CCP, similar to the proposal for its use within CFP.

What are the advantages of this modification? January 2001 What are the advantages of this modification? CFP provides guaranteed transmission opportunities while CCP within CFP provides flexibility of distributed access. Best of both worlds! Little needs to be done at ESTA for priority differentiation within CCP No new back-off mechanism at ESTA No issues due to legacy stations Bandwidth wastage within CFP is eliminated RR frames, TCSIZE indications can be used in CCP also to continue to provide information to EAP/EPC EAP/EPC can choose CCP for ONE priority and hence completely avoid the collision among frames of different priorities ESTAs can use CP (also) for control/management/non-Qos data transfers. Any proposed EDCF mechanism works better in CCP (than in CP)

Resolving issues in 802.11-1999 to assist in Qos operation January 2001 Resolving issues in 802.11-1999 to assist in Qos operation

January 2001 What are the issues? There are five issues in 802.11-1999 that needs to be fixed for proper Qos operation. The issues are, Not limiting transmission of all STAs during CP to end before the start of CFP for all PHYs - Should be fixed in clause 9, Qos baseline document (#360). Not limiting transmission of a polled STA to a duration as decided by PC - Should be fixed in clause 9, Qos baseline document (#360). Not specifying Flow spec - Fixed in doc#360, clause 7. But needs modifications Limiting the maximum CFP duration to accommodate at least MaxMPDUTime in CP - modification suggested later in the presentation. The polling list clause is outdated and is an hindrance to the Qos operation - must be updated appropriately and made into an informal or guidance section in an appendix

What are the modifications? (1) January 2001 What are the modifications? (1) Not limiting transmission of all STAs during CP to end before the start of CFP for all PHYs - Should be fixed in clause 9, doc#360. Not limiting transmission of a polled STA to a duration as decided by PC - Should be fixed in clause 9, doc#360

What are the modifications? (2) January 2001 What are the modifications? (2) Removing the limitation on the maximum CFP duration to accommodate at least MaxMPDUTime in CP Legacy STAs extend the CP by at least one DCF frame exchange time Hence this limit only adds to the problem even when all the ESTAs are operating in CFP Let EAP/EPC decide the duration of CFP and CP independent of this limit depending solely on amount of Qos-traffic and non-Qos traffic expected in both CFP and CP. This limit must be modified to minimum size frame time. If a legacy station extends its transmission into CFP, the time taken for the frame exchange automatically makes the MaxMPDUTime limit.

What are the modifications? (3) January 2001 What are the modifications? (3) Removing normative text on polling list The current polling list is an outdated and an hindrance to provide good Qos based on decisions made at EPC Polling list must take into account the priorities of TC and number of such priority TCs from each ESTA. According to the currently specified list processing rules, ALL (CF-pollable) STAs must be polled before polling the same STA again. If an (E)STA has more data to send while others don’t, this results in lot of bandwidth wastage in trying to poll all the (E)STAs before returning to the (E)STA with data. According to the currently specified list processing rules, the polls must be done in the ascending order of AID. Given that an high priority traffic might start at any time during the QBSS life time, this is not applicable anymore.

What are the modifications? (4) January 2001 What are the modifications? (4) The proposed Qos mechanisms allow sophisticated schedulers to be used. Hence it is proposed that the polling list creation, updation and processing clause (9.3) be appropriately modified and made informal or guideline material in an appendix rather than a normative clause.

What are the modifications? (5) January 2001 What are the modifications? (5) Making Flow-spec exchange mandatory for all traffic Flow spec is an important piece of information that EAP/EPC can make use of during channel time allocation Depending on the type of higher layer protocol used, this information may or may not be available at EAP/EPC. Hence it is proposed that The generic_action frame with flow spec be allowed to be sent by all ESTAs with a TC connection. The EAP/EPC makes use of this information for allocation of Tx-ops This merges the level-3 concept of parameterization of flows with the level-2 concept of prioritized channel allocation mechanism. And hence avoids one another level of Qos in 802.11e MAC.

January 2001 Resolving issues in Qos-Tx-ops in the current Qos-baseline document - The Qos-Tx-op is defined as the Tx-op allocated using Qos-CF-poll, CF-multi-poll and CF-Sch frames

January 2001 What are the issues? There are five issues in Tx-op operation that needs to be fixed for proper Qos. The issues are related to, Definition of non-final bit is ambiguous lack of retry procedure lack of procedure for EPC to take the left over Tx-op Lack of indication of time for which data is pending at an ESTA Use of Duration/ID field and use of CF-Ack Lack of clarification on use of container frame Use of sequence control for Qos-Null-data (no data) frames

What are the modifications? (1) January 2001 What are the modifications? (1) Remove Non-final bit from the header Original proposal of Non-final bit was for scheduled transmissions and for the next ESTA in line for transmission to take over the channel the generalization of use of this bit causes ambiguity and adds unnecessary complexity when the transmission opportunities are not scheduled The problem gets worse when the frame containing Non-final-bit is not correctly received by the EAP/EPC Hence it is proposed to remove the Non-final bit and let the EAP/EPC make use of the left over Qos-Tx-op as described later in this paper.

What are the modifications? (2) January 2001 What are the modifications? (2) Remove Override bit from the header Introduces the complexity at ESTA that is having to change its queuing dynamically based on this bit If the frame containing the Override bit is not correctly received at the ESTA in question, this can cause confusion and hence lead to collisions during CFP Given that the EAP/EPC needs to plan the Qos-Tx-ops due to CF-Sch and CF-multi-poll ahead of time, the benefit of this is not evident. Hence it is proposed to remove the override bit from the new header definition

What are the modifications? (3) January 2001 What are the modifications? (3) Procedure for retries during Qos-Tx-op ESTAs must use PIFS to retry their frames, if required, during their own Qos-Tx-op EAP/EPC must use DIFS (yes, it is DIFS!) to take over the channel when there is left over Qos-Tx-op time. The owner of Qos-Tx-op must recognize the channel getting busy after DIFS and must interpret that as the end of its Qos-Tx-op Since EAP/EPC uses DIFS and the owner of Qos-Tx-op uses PIFS for retries, the channel access is contention free

What are the modifications? (4) January 2001 What are the modifications? (4) Rules for using the left over Qos-Tx-op EAP/EPC must use DIFS to take over the channel when there is left over Qos-Tx-op time. EAP/EPC can use the left over time in the current Qos-Tx-op for any time bound frame exchange as long as such an exchange does not cause any other already allocated Qos-Tx-ops to be violated. An exception to this is EAP/EPC sending CF-Sch to terminate all the allocated scheduled Qos-Tx-ops Or reallocating unallocated time within CFP interval(s). EAP/EPC must NOT nest CF-multi-polls (especially with second CF-multi-poll using left over Qos-Tx-op) that causes two Qos-Tx-ops for the same ESTA.

What are the modifications? (5) January 2001 What are the modifications? (5) Indication of “max elapsed time” for the TC Each TC might have data pending for transmission for a while EAP/EPC must provide preference to TC with the longest waiting frame for a given priority when allocating Tx-ops Hence both RR frames and the Qos-MAC-Header must contain a field to indicate the “max elapsed time” to EAP/EPC

What are the modifications? (6) January 2001 What are the modifications? (6) Use Duration/ID field for Duration value in Qos-Data within CFP too Within Qos-Tx-op the ESTAs must use Duration/ID field to indicate the duration till the end of their current Qos-Tx-op. Similar proposal has already been made for enhanced CC frame and transmissions during CCP. Move the “special” contents of the Duration/ID field in Qos baseline proposal to a new field (along with the “max elapsed time” indication) as shown in the next slide

Duration covering the Tx-op January 2001 What are the modifications? (7) The Qos-DATA Frame format Duration covering the Tx-op Modified TC ID field: Encoding of each field is described in the next slide

What are the modifications? (8) January 2001 What are the modifications? (8) Restrict the response frames to fit within the Qos-Tx-op of the ESTA that is initiating the frame exchange. Since the “Dur/ID” field provides the information on end of Qos-Tx-op, the responding ESTA must be able to limit its transmission so as not to exceed the current Tx-op. However, it is unfair for one ESTA to take another’s Qos-Tx-op. The owner of Qos-Tx-op or EPC will not be able to estimate the time required in the Qos-Tx-op properly! Hence there has to be a bit in the data frames allowing or not allowing CF-Ack. The “Ack-policy” combination of ‘11’ (binary) for this purpose. That is, when the Ack-policy=‘11’, the responding station sends the control ‘ack’ frame as the response and does not send CF-Ack with its own data.

What are the modifications? (9) January 2001 What are the modifications? (9) Ack policy, Tx-op-limit and priority remain as it is defined in 360r1 TC-SIZE is encoded in a non-linear fashion TC-SIZE represents the time (not bits or bytes) as the only ESTA knows the PHY rate at which those bits are transmitted TC-SIZE encoding is as follows: values of [7-0] represents the requested time in the unit of 256 microseconds. Hence [0-1792] microseconds is represented values of [8-15] represents the requested time in the unit of 384 microseconds. Hence [3072-5760] microseconds is represented values of [16-31] represents the requested time in the unit of 512 microseconds. Hence [8192-15872] microseconds is represented values of [32-63] represents the requested time in the unit of 1024 microseconds (TU). Hence [32-63] TUs is represented Max-Elapsed-Time is encoded as TUs (0-32 TUs)

What are the modifications? (10) January 2001 What are the modifications? (10) Container frame must be treated like any other management frame Normal-ack required. Ack can be avoided if the container frame is addressed to a group (multicast or broadcast), even though the frames may belong to only one ESTA. No TCID field Use sequence control from the same counter that is used by other management frames Note that since management frames can transmitted in any priority range in the proposed CCP, container frames are not held back.

What are the modifications? (11) January 2001 What are the modifications? (11) Ignore Sequence control in Qos-Null-data (no data) frames Qos-data frames use sequence control that is specific to each (SA, DA, priority) triplet. Qos-Null-data frames are dynamically constructed by the low-level entity that needs to be in the hardware for SIFS response time Hence the sequence control for Qos-Null-data frames must be dynamically assigned and the succeeding frames in the queue are dynamically adjusted for their sequence control. OR all the sequence control counters must be maintained in hardware. This is an implementation nightmare! At minimum, Qos-Null-data frames must use the same counter as the legacy frames to ease the implementation burden. But since the Qos-Null-data frames are not used in duplication detection and/or the delayed-ack mechanisms, the sequence control for these frames can as well be ignored.