Packet Dispersion in IEEE 802 11 Wireless Networks

  • Slides: 37
Download presentation
Packet Dispersion in IEEE 802. 11 Wireless Networks Mingzhe Li, Mark Claypool and Bob

Packet Dispersion in IEEE 802. 11 Wireless Networks Mingzhe Li, Mark Claypool and Bob Kinicki WPI Computer Science Department Worcester, MA 01609 rek@cs. wpi. edu Second International Workshop on Performance and Management of Wireless Local Area Networks (P 2 MNet) Tampa, Florida, November 14, 2006

Outline • • • 2 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model

Outline • • • 2 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model Validation Model Analysis Conclusions P 2 MNet Workshop November 14, 2006

Introduction • Bandwidth estimation techniques focus on network capacity or available bandwidth. • Most

Introduction • Bandwidth estimation techniques focus on network capacity or available bandwidth. • Most bandwidth estimation has involved only wired networks. • This paper focuses on packet dispersion in wireless LANs (WLANs). 3 P 2 MNet Workshop November 14, 2006

Outline • • • 4 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model

Outline • • • 4 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model Validation Model Analysis Conclusions P 2 MNet Workshop November 14, 2006

Packet Dispersion Techniques • Packet pair dispersion – two equal-sized packets are sent backto-back

Packet Dispersion Techniques • Packet pair dispersion – two equal-sized packets are sent backto-back through the network. • Packet train dispersion – multiple back-to-back probe packets are sent through the network. 5 P 2 MNet Workshop November 14, 2006

Packet Pair Dispersion narrow link 6 P 2 MNet Workshop November 14, 2006

Packet Pair Dispersion narrow link 6 P 2 MNet Workshop November 14, 2006

NS-2 Additions to Model WLANs • Receiver Based Auto Rate (RBAR) module was re-implemented

NS-2 Additions to Model WLANs • Receiver Based Auto Rate (RBAR) module was re-implemented in NS 2. 27 to model 802. 11 dynamic rate adaptation [ Sadeghi et al. ] • An NS-2 extension to model Rayleigh fading was implemented [Ratish et al. ] 7 P 2 MNet Workshop November 14, 2006

Wireless Rate Adaptation under Rayleigh Fading NS-2 simulation results 8 P 2 MNet Workshop

Wireless Rate Adaptation under Rayleigh Fading NS-2 simulation results 8 P 2 MNet Workshop November 14, 2006

Wireless Traffic Classification • Probing traffic – Packet pairs or trains sent downstream through

Wireless Traffic Classification • Probing traffic – Packet pairs or trains sent downstream through the AP to the wireless client. • Crossing traffic – Downstream traffic going through the AP to other wireless clients. • Contending traffic – Upstream traffic that accesses the shard wireless channel and competes with probe packets on the estimated path. 9 P 2 MNet Workshop November 14, 2006

Wireless Traffic Classification 10 P 2 MNet Workshop November 14, 2006

Wireless Traffic Classification 10 P 2 MNet Workshop November 14, 2006

WLAN Packet Dispersion Issues • Wireless frame retries – increase the packet delay variance

WLAN Packet Dispersion Issues • Wireless frame retries – increase the packet delay variance that produce packet dispersion inconsistencies. • • 11 Dynamic Rate Adaptation Fading channel Noisy wireless channel - BER Contending Traffic P 2 MNet Workshop November 14, 2006

Simulated Wireless Effects on Bandwidth Estimation • CDF represents 1000 packet-pair estimates • 1000

Simulated Wireless Effects on Bandwidth Estimation • CDF represents 1000 packet-pair estimates • 1000 -byte probe packets and CBR contending packets • Contending traffic = 1 Mbps upstream flow 12 P 2 MNet Workshop November 14, 2006

Outline • • • 13 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model

Outline • • • 13 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model Validation Model Analysis Conclusions P 2 MNet Workshop November 14, 2006

WLAN Packet Dispersion Model Assumptions Bottleneck link is the WLAN last hop. No crossing

WLAN Packet Dispersion Model Assumptions Bottleneck link is the WLAN last hop. No crossing traffic. • Dispersion, T, between two packets in a packet pair is characterized in terms of E[T] and V[T] for a WLAN that includes: packet size, link data rate, BER and access methods (base or RTS/CTS). • While wireless channel conditions can be characterized by received signal strength indicator (RSSI), SNR and BER, our simplified model assumes these other factors impact BER. – The impact of channel conditions on bandwidth estimation is evaluated via V[T], the packet dispersion variance. 14 P 2 MNet Workshop November 14, 2006

WLAN Packet Dispersion Model • From previous 802. 11 models, we build a new

WLAN Packet Dispersion Model • From previous 802. 11 models, we build a new model for wireless packet dispersion where dispersion is defined as the delay between the arrival times of the first and second packets in the packet pair. • Thus, the model must include the delay before the transmission of the second packet, E[D], and the time to transmit it, Ts, and dispersion is: E[T] = E[D] + Ts 15 P 2 MNet Workshop November 14, 2006

WLAN Packet Dispersion Model • Since E[D] depends on the wireless link rate, Cl

WLAN Packet Dispersion Model • Since E[D] depends on the wireless link rate, Cl , the average packet size, L and the number of nodes in the contention domain, n, we have E[D] = d (Cl, L, n) • Similarly, for the transmission time, Ts = ts (Cl, L) 16 P 2 MNet Workshop November 14, 2006

Bandwidth Estimate • Defining, Cest, as the bandwidth estimate based on the wireless packet

Bandwidth Estimate • Defining, Cest, as the bandwidth estimate based on the wireless packet pair dispersion model, our model is defined by: 17 P 2 MNet Workshop November 14, 2006

Outline • • • 18 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model

Outline • • • 18 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model Validation Model Analysis Conclusions P 2 MNet Workshop November 14, 2006

Model Validation • Validation results include an ideal WLAN and a scenario with contention

Model Validation • Validation results include an ideal WLAN and a scenario with contention and BER. • Ideal validation includes NS-2 simulations, analytic model and actual measurements. 19 P 2 MNet Workshop November 14, 2006

Ideal WLAN Scenario The ideal scenario consists of an AP with a single wireless

Ideal WLAN Scenario The ideal scenario consists of an AP with a single wireless client for both basic (BAS) and RTS/CTS access methods. Simulations: 500 pairs Measurements: 100 pairs 20 P 2 MNet Workshop November 14, 2006

Model Validation Random Simulation Topology 21 Model Parameters P 2 MNet Workshop November 14,

Model Validation Random Simulation Topology 21 Model Parameters P 2 MNet Workshop November 14, 2006

Models with Contention and Wireless Transmission Errors in the Bandwidth Estimation Model Compared with

Models with Contention and Wireless Transmission Errors in the Bandwidth Estimation Model Compared with Simulations Error Free 22 BER = 10 -5 RTS/CTS Basic Mean Error 8. 05% 4. 90% 9. 40% 7. 67% Standard Deviation 6. 72% 4. 28% 5. 30% 3. 82% P 2 MNet Workshop November 14, 2006

Outline • • • 23 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model

Outline • • • 23 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model Validation Model Analysis Conclusions P 2 MNet Workshop November 14, 2006

Model Analysis • In a non-saturated WLAN with low BER, low contention, packet-pair dispersion

Model Analysis • In a non-saturated WLAN with low BER, low contention, packet-pair dispersion estimates the maximal channel capacity, effective capacity, Ce. • With considerable contending traffic and/or BER, the dispersion metric is referred to as the achievable throughput, At , for the current level of contending traffic. 24 P 2 MNet Workshop November 14, 2006

Model Analysis • Effective Capacity • Achievable Bandwidth • Relationship with Available Bandwidth 25

Model Analysis • Effective Capacity • Achievable Bandwidth • Relationship with Available Bandwidth 25 P 2 MNet Workshop November 14, 2006

Achievable Throughput 26 P 2 MNet Workshop November 14, 2006

Achievable Throughput 26 P 2 MNet Workshop November 14, 2006

Impact of Channel Rate on Effective Capacity 27 P 2 MNet Workshop November 14,

Impact of Channel Rate on Effective Capacity 27 P 2 MNet Workshop November 14, 2006

Impact of BER on Achievable Throughput Five node, BAS wireless network with 1500 -byte

Impact of BER on Achievable Throughput Five node, BAS wireless network with 1500 -byte packets no achievable throughput 28 P 2 MNet Workshop November 14, 2006

Impact of RTS/CTS on Achievable Throughput Model Results for Wireless Networks with 1500 Byte

Impact of RTS/CTS on Achievable Throughput Model Results for Wireless Networks with 1500 Byte Packets 29 P 2 MNet Workshop November 14, 2006

Standard Deviation of Achievable Throughput 30 P 2 MNet Workshop November 14, 2006

Standard Deviation of Achievable Throughput 30 P 2 MNet Workshop November 14, 2006

Packet Size Effect on Standard Deviation of Bandwidth Estimations Five node, BAS wireless network

Packet Size Effect on Standard Deviation of Bandwidth Estimations Five node, BAS wireless network with 1500 -byte packets 31 P 2 MNet Workshop November 14, 2006

Impact of Channel Rate on Standard Deviation of Bandwidth Estimations 32 P 2 MNet

Impact of Channel Rate on Standard Deviation of Bandwidth Estimations 32 P 2 MNet Workshop November 14, 2006

Impact of BER on Standard Deviation of Bandwidth Estimations 33 P 2 MNet Workshop

Impact of BER on Standard Deviation of Bandwidth Estimations 33 P 2 MNet Workshop November 14, 2006

Outline • • • 34 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model

Outline • • • 34 Introduction Packet Dispersion in WLANs WLAN Packet Dispersion Model Validation Model Analysis Conclusions P 2 MNet Workshop November 14, 2006

Conclusions • Packet dispersion measures the effective capacity and the achievable throughput of a

Conclusions • Packet dispersion measures the effective capacity and the achievable throughput of a wireless network instead of the capacity as in a wired network. • Wireless channel conditions, such as BER and RTS/CTS access method impact the bandwidth estimation results and the variance of the results. 35 P 2 MNet Workshop November 14, 2006

Thank You !! Packet Dispersion in IEEE 802. 11 Wireless Networks Mingzhe Li, Mark

Thank You !! Packet Dispersion in IEEE 802. 11 Wireless Networks Mingzhe Li, Mark Claypool and Bob Kinicki WPI Computer Science Department Worcester, MA 01609 rek@cs. wpi. edu http: //web. cs. wpi. edu/~rek/ Second International Workshop on Performance and Management of Wireless Local Area Networks (P 2 MNet) Tampa, Florida, November 14, 2006

37 P 2 MNet Workshop November 14, 2006

37 P 2 MNet Workshop November 14, 2006