Presentation is loading. Please wait.

Presentation is loading. Please wait.

QoS of Voice over 802.11 with NS simulator Prepared by: Yoshpa Benny Shraer Alexander Vainer Albert Instructors: Prof. Reuven Cohen Mr. Itai Dabran.

Similar presentations


Presentation on theme: "QoS of Voice over 802.11 with NS simulator Prepared by: Yoshpa Benny Shraer Alexander Vainer Albert Instructors: Prof. Reuven Cohen Mr. Itai Dabran."— Presentation transcript:

1 QoS of Voice over 802.11 with NS simulator Prepared by: Yoshpa Benny Shraer Alexander Vainer Albert Instructors: Prof. Reuven Cohen Mr. Itai Dabran

2 Overview 802.11 - applies to wireless LANs and provides 1 or 2 Mbps transmission in the 2.4 GHz band using either frequency hopping spread spectrum (FHSS) or direct sequence spread spectrum (DSSS). Problem: No QoS 802.11a - an extension to 802.11 that provides theoretically up to 54 Mbps in the 5GHz band, but realistically achieves 20-25 Mbps under normal conditions. 802.11a uses an orthogonal frequency division multiplexing encoding scheme (OFDM) 802.11b (802.11 High Rate or Wi-Fi) - an extension to 802.11 that provides 11 Mbps transmission in the 2.4 GHz band. 802.11b uses only DSSS. 802.11b was a 1999 ratification to the original 802.11 standard, allowing wireless functionality comparable to Ethernet.

3 Overview Both 802.11a and 802.11b have two channel accessing mechanisms: PCF - Point Coordination Function Based on polling technique – each station is polled in turn and stations with a packet pending for transmission sends the packet upon being polled. A dynamic list of stations may be maintained by the AP in order to increase efficiency. DCF - Distributed Coordination Function Based on CSMA/CA – stations contend for the channel. Two basic schemes are used: – Two way handshake: Acknowledgement is sent by the receiver to the sender upon successful reception of a packet. The acknowledgement is needed since the sender cannot determine whether its transmission was successful only by listening to it. – Four way handshake: RTS/CTS mechanism, and then proceed with ACKs as above.

4 DCF – An Example If the first station senses channel idle for DIFS it sends a DATA packet. Otherwise it waits for the channel to be idle for DIFS and than selects random backoff – like the second source station above. Backoff is selected from the range [0, CW-1]. At start, CW = CWmin. After collision, CWnew = min {CWold * 2, CWmax}. CW is reset to CWmin after successful transmission Destination station senses the channel idle for SIFS and then sends ACK. NAV is used for virtual carrier sensing – info is sent in data packets to indicate how long the source intends to occupy the channel.

5 QoS 802.11a, 802.11b MAC don’t support service differentiation. Relevant parameters for service differentiation:  PF – Persistence Factor. Determines how to increase CW after collision.  Frame Size – For Voice packets it is determined by the voice codec used. QoS is very important for applications such as voice.  DIFS  CWmin  CWmax Explained in previous slide 802.11e – An extension of 802.11 designed to improve its medium access mechanism and to add support for service differentiation. Uses the HCF – hybrid coordination function which is queue based service differentiation scheme that uses both DCF and PCF enhancements - EDCF and EPCF.

6 Project Goals  Improvement of NS2 network simulator in order to simulate statistical QoS in 802.11.  Show that Uplink/Downlink problem exists in EDCF and suggest different solutions using QoS parameters.  Investigation of QoS for Voice streams in 802.11e, EDCF and comparison to 802.11b.

7 QoS Parameters Implemented  DIFS – In 802.11 DIFS is defined as a function of SIFS and is the same for all mobile nodes. In EDCF DIFS is different for every service class (Voice/Video/Data/etc.).  CWmin and Persistence Factor (PF) – In 802.11 the same for all mobile nodes. In EDCF it is different for every service class. EDCF computes the CW differently then DCF: CWnew = ( (CWold + 1) * PF ) – 1 (up to CWmax)  The lower these parameters are, the higher priority the service - class gets in accessing the channel.

8  Investigate QoS indications (for voice packets), 1. Average Throughput (per voice connection). 2. Average Latency. 3. Packet Loss percentage.  A function of number of voice streams  A function of different QoS parameters (DIFS, CWmin, PF) Goal 1  Show improvement of our QoS implementation over the regular 802.11 MAC

9 Simulated Network  Actually, all wireline network is not simulated, since it is not interesting for our goal. Instead we let the access point to produce packets on all connections that go from wireline to wireless.

10 Setting Connections  There are N + 4 connections: half from a wireless node to a wireline node and half the other way (the connections are in one-direction)  Each wireless station has a different wireline station it “talks” to.  Voice Connections: - CBR over UDP. - Packet size (with all overhead) : 180 bytes as in G711 voice codec - Packet inter-arrival time – 20ms  Best Effort Connections: DATA - FTP over TCP. - Packet size (with all overhead) : 1560 bytes  There are N voice connections. The parameters we investigate are taken only from these connections

11 Parameters: Voice: DIFS - 0.000020, CWMin - 15, PF - 2 Parameters: Voice: DIFS - 0.000030, CWMin - 31,PF - 2 Parameters: Voice: DIFS - 0.000040,CWMin - 3, PF - 2 Average Throughput as function of VOIP Calls ----- 802.11e ----- 802.11b Other Parameters: Data: DIFS - 0.000080, CWMin - 63, PF – 2 CWMin of 802.11b – 63 CWMax – 1023

12 Average cbr Latency as function of VOIP Calls Parameters: Voice: DIFS - 0.000020,CWMin - 15,PF - 2 Parameters: Voice: DIFS - 0.000030,CWMin - 31,PF - 2 Parameters: Voice: DIFS - 0.000040,CWMin - 3,PF - 2 ----- 802.11e ----- 802.11b Other Parameters: Data: DIFS - 0.000080, CWMin - 63, PF – 2 CWMin of 802.11b – 63 CWMax – 1023

13 Percentage of dropped cbr packets as function of VOIP Calls Parameters: Voice: DIFS - 0.000020,CWMin - 15,PF - 2 Parameters: Voice: DIFS - 0.000030,CWMin - 31,PF - 2 Parameters: Voice: DIFS - 0.000040,CWMin - 3, PF - 2 ----- 802.11e ----- 802.11b Other Parameters: Data: DIFS - 0.000080, CWMin - 63, PF – 2 CWMin of 802.11b – 63 CWMax – 1023

14 Conclusions  Our changes do work!  The improvement of service to Voice could be clearly seen when number of voice connections is relatively small (up to 3 times more than number of background connections).  The Voice class, to which we gave lower parameters (higher priority) got better service than in 802.11b => QoS  When the number of voice connection becomes large, the performance of 802.11b and 802.11e (EDCF) is very much alike, since giving priority has meaning only when there are many connections with lower priority. If there are many voice connections, they compete mainly among them and not so much with the Data connections.  Channel capacity limits the number of possible connections through the channel. The channel contention method doesn’t help either.

15 Uplink / Downlink Problem  EPCF (Polling) – The average throughput is the same for uplink and downlink connection, since the AP gets access to the channel at least as often as any other station (its up to the AP to decide how often). Downlink – Link from AP to the mobile station Uplink – Link from the mobile station to the AP.  EDCF (CSMA/CA) – The AP is heavily loaded with traffic from the wireline network side and it needs to contend the channel with other mobile stations. This will introduce larger uplink throughput then the average downlink throughput.

16 Show that Uplink Downlink problem exists in EDCF. Goal 2 Check dependence on number of mobile nodes Suggest solutions using QoS parameters: – DIFS – CWmin – Persistence Factor

17 Uplink/Downlink Throughput as function of DIFS Other Parameters: CWMin of AP = 15, PF of AP = 2 DIFS of Voice = 50,CWMin of Voice = 31, PF of Voice = 2 DIFS of BG = 80,CWMin of BG = 63, PF of BG = 2 CWMax – 1023 ----- Uplink Throughput ----- Downlink Throughput Num of Stations = 8Num of Stations = 16 Num of Stations = 26

18 Uplink/Downlink Throughput as function of CWMin ----- Uplink Throughput ----- Downlink Throughput Other Parameters: DIFS of AP = 40, PF of AP = 2 DIFS of Voice = 50,CWMin of Voice = 31, PF of Voice = 2 DIFS of BG = 80,CWMin of BG = 63, PF of BG = 2 CWMax – 1023 Num of Stations = 8Num of Stations = 16 Num of Stations = 26

19 Uplink/Downlink Throughput as function of PF Num of Stations = 8 Num of Stations = 16 Num of Stations = 26 ----- Uplink Throughput ----- Downlink Throughput Other Parameters: DIFS of AP =40,CWMin of AP = 15, DIFS of Voice = 50,CWMin of Voice = 31, PF of Voice = 2 DIFS of BG = 80,CWMin of BG = 63, PF of BG = 2 CWMax – 1023

20 Conclusions – cont. dynamically  Access Point should dynamically adjust its medium access parameters (DIFS, CWmin and PF) depending on number of mobile stations and network load. It should aspire to get at least equal service as all the mobile stations it serves (together).  AP should also consider the type of data (priority classes) that passes through it when adjusting its own medium access parameters.  When adjusting AP’s priority, it is better to use DIFS and CWMin rather than PF, and when PF is used, it should be accompanied with changes in CWMin, since sometimes the changes PF inflicts are too strong and not very predictable unlike the other two parameters.  Each one of the three parameters investigated allows to overcome the uplink/downlink problem described earlier.  We have shown that as the number of stations (and thus network load) increases, the AP has to get higher priority (lower QoS parameters) to reach the uplink = downlink point.

21 Bibliography IEEE 802.11e Wireless LAN for Quality of Service Scheduling of Voice Packets in a Low-Bandwidth Shared Medium Access Network, Reuven Cohen, Liran Katzir (Technion, Israel) Achieving Service Differentiation and High Utilization in 802.11 Vasilios A. Siris Supporting VBR VoIP Traffic in IEEE 802.11 WLAN in PCF Mode Dongyan Chen, Sachin Garg, Martin Kappes and Kishor S. Trivedi Supporting VoIP Traffic in IEEE 802.11 WLAN with Enhanced MAC for QoS - Dongyan Chen, Sachin Garg, Martin Kappes and Kishor S. Trivedi


Download ppt "QoS of Voice over 802.11 with NS simulator Prepared by: Yoshpa Benny Shraer Alexander Vainer Albert Instructors: Prof. Reuven Cohen Mr. Itai Dabran."

Similar presentations


Ads by Google