HCF and EDCF Simulations

Slides:



Advertisements
Similar presentations
Doc.: IEEE /413r0 Submission S. Choi, Philips Research July 2001 Slide 1 Can EDCF Support QoS? Sunghyun Choi Philips Research-USA Briarcliff Manor,
Advertisements

Doc.:IEEE /525Ar0 Submission September 2002 Mathilde Benveniste, Avaya Labs Slide 1 Simplifying Polling Mathilde Benveniste
Achieving Quality of Service in Wireless Networks A simulation comparison of MAC layer protocols. CS444N Presentation By: Priyank Garg Rushabh Doshi.
Doc.: IEEE /133 Submission March 2001 G. Chesson, A. Singa - Atheros Slide 1 VDCF Simulations Greg Chesson, Aman Singla,
Presented by Scott Kristjanson CMPT-820 Multimedia Systems Instructor: Dr. Mohamed Hefeeda 1 Cross-Layer Wireless Multimedia.
1 Solutions to Performance Problems in VOIP over Wireless LAN Wei Wang, Soung C. Liew Presented By Syed Zaidi.
1 QoS Schemes for IEEE Wireless LAN – An Evaluation by Anders Lindgren, Andreas Almquist and Olov Schelen Presented by Tony Sung, 10 th Feburary.
QoS of Voice over with NS simulator Prepared by: Yoshpa Benny Shraer Alexander Vainer Albert Instructors: Prof. Reuven Cohen Mr. Itai Dabran.
Voice Traffic Performance over Wireless LAN using the Point Coordination Function Wei Supervisor: Prof. Sven-Gustav Häggman Instructor: Researcher Michael.
Providing QoS in Ad Hoc Networks with Distributed Resource Reservation IEEE802.11e and extensions Ulf Körner and Ali Hamidian.
Company LOGO Provision of Multimedia Services in based Networks Colin Roby CMSC 681 Fall 2007.
Voice over WiFi R 張素熒 R 朱原陞 R 王振宇
Methods for providing Quality of Service in WLANs W.Burakowski, A. Beben, J.Sliwinski Institute of Telecommunications, Warsaw University of Technology,
Service differentiation mechanisms for IEEE based wireless networks § Srikant Kuppa & Ravi Prakash Distributed Systems Laboratory The University.
Qos support and adaptive video. QoS support in ad hoc networks MAC layer techniques: – e - alternation of contention based and contention free periods;
Doc.: IEEE /601r0 Submission Harada Yasuo, Matsushita Electric Ind. Slide 1 November20 01 Delayed Acknowledgement v.s. Normal Acknowledgement.
An Energy Efficient MAC Protocol for Wireless LANs, E.-S. Jung and N.H. Vaidya, INFOCOM 2002, June 2002 吳豐州.
Submission doc.: IEEE 11-14/0866r0 July 2014 Johan Söder, Ericsson ABSlide 1 Traffic modeling and system capacity performance measure Date:
Quality of Service Schemes for IEEE Wireless LANs-An Evaluation 主講人 : 黃政偉.
1 A Cross-Layer Scheduling Algorithm With QoS Support in Wireless Networks Qingwen Liu, Student Member, IEEE, Xin Wang, Member, IEEE, and Georgios B. Giannakis,
Doc.: IEEE /457 Submission December 2000 Mathilde Benveniste, AT&T Labs - ResearchSlide 1 An Enhanced-DCF Proposal Based on ‘Tiered Contention’
doc.: IEEE /409r0 Submission March 2002 Mathilde BenvenisteSlide 1 Persistence Factors in EDCF Mathilde Benveniste
Performance Comparison of Ad Hoc Network Routing Protocols Presented by Venkata Suresh Tamminiedi Computer Science Department Georgia State University.
AIFS – Revisited Mathilde Benveniste
QoS & Queuing Theory CS352.
Balancing Uplink and Downlink Delay of VoIP Traffic in WLANs
IEEE e Performance Evaluation
Topics discussed in this section:
Delayed Acknowledgement v.s. Normal Acknowledgement
Overview of ‘Tiered Contention’ Multiple Access (TCMA)
Empirically Characterizing the Buffer Behaviour of Real Devices
An Access Mechanism for Periodic Contention-Free Sessions
Wireless Mesh Networks
IEEE : Wireless LANs ALOHA, Slotted ALOHA
September 2005 Performance Evaluation of the CCC MMAC Protocol for s Mesh Networks Date: Authors: Notice: This document has been prepared.
HCF medium access rules
Using Dynamic PCF to improve the capacity of VoIP traffic in IEEE 802
Quantification of Capture Effect in Near-Far Scenarios
EDCF TXOP Bursting Simulation Results
AIFS – Revisited Mathilde Benveniste
Provision of Multimedia Services in based Networks
Speaker:Fu-Yuan Chuang Advisor:Ho-Ting Wu Date:
Scheduling Algorithms in Broad-Band Wireless Networks
Simulation Results for QoS, pDCF, VDCF, Backoff/Retry
Multipoll, FEC, Persistence, Portals
Simulation for EDCF Enhancement Comparison
Texas Instruments Incorporated
EDCA Backoff Rules Mathilde Benveniste
VDCF Presentation Greg Chesson,
HCF Channel Access And Inter-BSS Channel Sharing
HCF medium access rules
Is the MAC sufficient for wireless high speed mesh LANs?
Delayed Acknowledgement v.s. Normal Acknowledgement
DL MU MIMO Error Handling and Simulation Results
HCF medium access rules
Student : Min-Hua Yang Advisor : Ho-Ting Wu Date :
Acknowledgement for Multicast Streams
A Fair Scheduling Scheme for HCF
TGe Metrics & Criteria Ad Hoc Group Summary Report
Delayed Acknowledgement v.s. Normal Acknowledgement
HCF medium access rules
Enhanced-DCF Wireless MAC Protocol: Some Simulation Results
Month 2000 doc.: IEEE /xxx July 2002
40 MHz Vs 20 MHz for video Date: Authors: July 2009
HCF Channel Access And Inter-BSS Channel Sharing
How MAC interacts with Capacity of Ad-hoc Networks – Interference problem Capacity of Wireless Networks – Part Page 1.
Infocom 2004 Speaker : Bo-Chun Wang
‘Shield’: Protecting High-Priority Channel Access Attempts
Satellite Packet Communications A UNIT -V Satellite Packet Communications.
Wireless MAC Multimedia Extensions Albert Banchs, Witold Pokorski
Presentation transcript:

HCF and EDCF Simulations October 2001 October 2001 HCF and EDCF Simulations Aman Singla Greg Chesson Atheros Communications, Inc. Aman Singla, Atheros

Overview Simulations Analysis October 2001 Overview Simulations First simulation results for both contention-based (EDCF) and polling-based (HCF) channel access methods Study the performance and efficiency of the two methods under the same scenarios Analysis Understand the operational behavior of EDCF and HCF Summary and Implications Aman Singla, Atheros

October 2001 Simulations Aman Singla, Atheros

Background Down-link characteristics are equivalent October 2001 Background Down-link characteristics are equivalent EAP (HC) always has preferential access _ EAP can transmit at pifs for both EDCF / HCF Concentrate on the up-link differences EDCF: contention-based access HCF: contention-free polled access Concentrate on guaranteed service scenarios Admission Control for all QoS streams Uniform 10% PER for all frames Channel degrades rapidly above 10% PER to make PHY rate backoff more effective Aman Singla, Atheros

Background (continued) October 2001 Background (continued) Overload the channel with background traffic best-effort traffic is not admission controlled A realistic, non-trivial HCF scheduler Uses all available information Reacts within sifs to data arrival and queue state information Implemented various heuristics and used the best performing Report results for a 36Mb/s 11a PHY Statistical principles apply at all PHY rates Aman Singla, Atheros

Simulations Scenario I October 2001 Simulations Scenario I Study the effect of number of QoS stations on latency and jitter Scenario II Study the QoS capacity and efficiency of the two channel access methods Scenario III Telephony case study Aman Singla, Atheros

QoS Streams - Traffic Model October 2001 QoS Streams - Traffic Model CBR streams with the inter-packet arrival interval based on a normal distribution around a period variance = period/4 mean = period Aman Singla, Atheros

Simulations – I (constant load) October 2001 Simulations – I (constant load) Scenario PHY = 36 Mb/s (11a), 10% PER Background traffic 3 backlogged queues (ESTAs) @ 1500 Byte packets QoS traffic Fixed 16.384 Mb/s load @ 2048 Byte packets Experiment Vary the # of ESTAs applying the QoS load 2 streams (ESTAs) @ 2 ms period, 4/8/16 streams (ESTAs) @ 4/8/16 ms period, 32 streams (ESTAs) @ 32 ms period Objective Study effect of changing number of ESTAs on end-to-end latency and jitter for QoS traffic Aman Singla, Atheros

EDCF – Latency Distribution October 2001 EDCF – Latency Distribution 80% of packets have end-to-end latency <= 2.5ms Aman Singla, Atheros

HCF – Latency Distribution October 2001 HCF – Latency Distribution 2 streams 32 streams Aman Singla, Atheros

Conclusions – Simulations I October 2001 Conclusions – Simulations I EDCF is resilient to the number of stations applying the QoS load This constant QoS load study shows that collision rate is largely independent of the number of stations HCF scheduler phase contributes significantly to end-to-end latency Scheduler phase: timing difference between poll arrival and data arrival Aman Singla, Atheros

Simulations – II (varying load) October 2001 Simulations – II (varying load) Scenario PHY = 36 Mb/s (11a), 10% PER Background traffic 3 backlogged queues (ESTAs) @ 1500 Byte packets QoS traffic 4 streams (ESTAs) CWMin/Max for EDCF = 7/15 Experiment Vary the packet size and period (applied load) of the QoS traffic a) 4 x 2048 Bytes @ 3.5 ms b) 4 x 1500 Bytes @ 2.7 ms c) 4 x 1024 Bytes @ 2.1 ms d) 4 x 512 Bytes @ 1.4 ms e) 4 x 256 Bytes @ 1.1 ms f ) 4 x 128 Bytes @ 0.9 ms Objective Compare channel capacity and efficiency for EDCF/HCF at similar performance levels Aman Singla, Atheros

Latency Distribution (2048 Byte pkts) October 2001 Latency Distribution (2048 Byte pkts) Aman Singla, Atheros

Latency Distribution (1500 Byte pkts) October 2001 Latency Distribution (1500 Byte pkts) Aman Singla, Atheros

Latency Distribution (1024 Byte pkts) October 2001 Latency Distribution (1024 Byte pkts) Aman Singla, Atheros

Latency Distribution (512 Byte pkts) October 2001 Latency Distribution (512 Byte pkts) Aman Singla, Atheros

Latency Distribution (256 Byte pkts) October 2001 Latency Distribution (256 Byte pkts) Aman Singla, Atheros

Latency Distribution (128 Byte pkts) October 2001 Latency Distribution (128 Byte pkts) Aman Singla, Atheros

Efficiency October 2001 Efficiency is measured by the amount QoS load for both EDCF / HCF Efficiency is measured by the amount of background load supported Background load using EDCF Background load using HCF Aman Singla, Atheros

Conclusions – Simulations II October 2001 Conclusions – Simulations II EDCF and HCF demonstrate comparable latency and jitter performance EDCF and HCF support similar levels of background traffic (are equally efficient) in the presence of the same QoS load for these case studies EDCF and HCF have similar QoS capacity QoS capacity is the maximal QoS load that can be supported for a particular set of service guarantees Latency and jitter degradation for EDCF and HCF occur at essentially the same QoS loads - i.e. loads greater than the QoS load curve on Slide 19 The contention-free period for HCF on Slide 19 is 90+% Aman Singla, Atheros

Simulations – III (telephony case) October 2001 Simulations – III (telephony case) Scenario PHY = 36 Mb/s (11a), 10% PER Background traffic 3 backlogged queues (ESTAs) @ 1500 Byte packets QoS traffic 36 uni-directional (up-link) telephony streams Each stream = 120 Bytes @ 10ms CWMin/Max for EDCF = 7/15 Objective Case study comparing EDCF and HCF Aman Singla, Atheros

Latency Distribution (36 phones) October 2001 Latency Distribution (36 phones) EDCF delivers 4.71 Mb/s of background traffic x HCF uses 75% of channel time for CFP Delivers 4.77 Mb/s of background traffic Aman Singla, Atheros

Conclusions – Simulations III October 2001 Conclusions – Simulations III EDCF provides better latency and jitter in this case study EDCF and HCF show similar efficiency (surplus bandwidth for background traffic) within the parameters of this case study EDCF and HCF have similar QoS capacity for this experiment HCF’s CFP duty cycle is 75% Latency exceeds 10ms with more phones Latency increase is caused by contention for EDCF, or scheduler conflicts for HCF Aman Singla, Atheros

October 2001 Analysis Aman Singla, Atheros

Operational behavior for EDCF October 2001 October 2001 Operational behavior for EDCF AIFS isolates contention between TCs Traffic from one TC contends mostly against traffic within the same TC AIFS projects an image of a lightly loaded network to higher priority applications by Separating contention/arbitration for higher priority traffic from contention/arbitration for lower priority traffic Deferring contention/arbitration for lower priority traffic until after contention/arbitration for higher priority traffic Transmission Contention/arbitration for high priority access Contention for low priority access AIFS Aman Singla, Atheros

Operational behavior for EDCF (cont.) October 2001 Operational behavior for EDCF (cont.) Admission Control is used to control the QoS load Controlled load _ controlled contention Contention as a function of the number of stations Contention at excessive loads Contention at low loads Aman Singla, Atheros

Operational behavior for HCF polling October 2001 Operational behavior for HCF polling Performance depends on the scheduler Contention free channel access Contention exists instead in the scheduler Scheduled Polling Phase difference between poll generation and data arrival contributes to latency Error control Channel error recovery contributes to scheduler complexity and overhead Admission control is needed Controlled load _ feasible schedule Aman Singla, Atheros

Factors affecting latency October 2001 Factors affecting latency Packet size, PHY rate, etc. Channel arbitration or scheduling artifacts Background traffic Channel packet errors Per-packet latency Aman Singla, Atheros

Summary (I) Latency and jitter are a reality of the wireless network October 2001 Summary (I) Latency and jitter are a reality of the wireless network Access method is just one contributing factor Controlled access does not guarantee fixed or low latency Similar latency and jitter for EDCF and HCF Service guarantees will be probabilistic Service guarantees will need Tx and Rx buffers Simulation studies help determine buffer sizes Similar buffer requirements for EDCF and HCF CBR source TX fifo Wireless channel RX fifo CBR sink Aman Singla, Atheros

October 2001 Summary (II) Service guarantees can be provided only on some limited QoS load Admission Control is required for both EDCF and HCF EDCF and HCF both work well within the limits EDCF and HCF both fail when QoS load exceeds the limit The QoS load limits for both EDCF and HCF are substantially the same ~90% of maximal throughput may be used for QoS load TCP congestion control arrives at the same operating point Remaining channel may be used for background (best-effort) traffic with equal efficiency for both EDCF and HCF Aman Singla, Atheros