By: Naeem Khademi (GS20561) Supervisor: Prof. Dr. Mohamed Othman - - PowerPoint PPT Presentation

• By: Naeem Khademi (GS20561)
• Supervisor: Prof. Dr. Mohamed Othman
• Date/Time: 10 November 2009 9:00 AM
• Duration : 30 min
• Faculty Of Computer Sc. & Information Technology (FSKTM)
• University Putra Malaysia (UPM)

IEEE 802.11 WLAN technology is the prevalent technology used by wireless users. TCP is the dominant transport protocol in Internet. Distributed Coordination Function (DCF) defined in IEEE 802.11 MAC provides all WLAN stations equal opportunity to access the transmission medium. Cross-layer interaction of TCP with DCF leads to flow level unfairness in WLAN.

Two main types of unfairness occur due to the cross-layer interaction of TCP and DCF in 802.11 MAC protocol [1]. 1. Size-based unfairness in benefit of long-lived flows against short- lived flows. 2. Direction-based unfairness in benefit of upstream flows against downstream flows.

Several solutions have been proposed to solve each type of unfairness phenomenon [2-4] each of which only solves one aspect of unfairness (either size-based or direction-based) while may deteriorating another aspect. The closest work to our research is done by Wu et al. [2]. It proposes a queue management policy at Access Point (AP) called Least Attained Service scheduling (LAS). LAS always gives service to the job that has received the least service so far without prior knowledge of job size .

LAS is used by [2] to provide direction-based fairness for upstream and download flows. Threshold-based LAS (TLAS) is used by [2] to provide size-based fairness. The main idea of TLAS is to give the newly arriving flow service priority up to a certain threshold (e.g. 50 packets). Once the threshold is reached, FIFO scheduling is employed on this flow. The reason behind TLAS proposal is to prevent starvation of long-lived TCP flows.

One of the key challenges that IEEE 802.11 WLAN technology faces is the lack of a unique solution to solve two typical types of unfairness which are sized based unfairness among flows in benefit of long-lived flows as well as direction based unfairness in benefit of upstream flows. Most proposed solutions solve one of these issues while deteriorated another

• ne or at least unable to improve it to an acceptable level.

The main objectives of this research are three-fold:

1. We demonstrate and evaluate different types

• f

unfairness phenomenon (short-lived versus long-lived flows as well as upstream versus downstream flows) in IEEE 802.11 WLAN. 2. We propose a novel queue management policy called “Threshold based Least Attained Service scheduling – Selective Acknowledgment Filtering (TLAS-SAF)” to alleviate mentioned types of unfairness to an acceptable level 3. We evaluate the efficiency and validation of our scheme using various simulation scenarios.

 In the 802.11 WLAN, a bandwidth asymmetry exists between contending upload and download flows.  the MAC layer contention parameters are all equal for the AP and the stations.  If K stations and an AP are always contending for the access to the channel, each host ends up having approximately 1/(K + 1) share of the total transmit opportunities over a long time interval.  Results in K/(K +1) of the transmissions being in uplink, while only 1/(K +1) of the transmissions belong to the downlink flows. Thus,

• verall upstream transmit opportunities will be 𝐿 times more than

downstream.

 802.11 WLAN is half-duplex.  At AP downstream queue, data packet loss (downstream) leads to Triple-Duplicate ACK mechanism and halving Congestion Window.  ACK loss impact on flow throughput can be seen negligible relying on cumulative nature of TCP ACK.  When upstream ACKs fill the AP buffer, downstream data packets will be dropped. Several consecutive loss of data packets result in RTO which sets congestion window to one.  This leads to aggressive upstream flows and conservative downstream flows.

The minimum buffer size to achieve direction-based fairness among all flows is determined by Pilosof et al. [4] as where 𝜖 is number of ACK packet per data packets in TCP and 𝑋 is the TCP advertised receiver window for all flows and 𝑂 and 𝑁 refers to the number of downstream and upstream flows respectively.

 TCP congestion and flow control mechanism is designed in such a way that large data flows can enjoy from the best effort bandwidth allocation.  Studies have revealed that Internet traffic shows a high variability property and most of the TCP flows are short (e.g. popular web transfer), while more than half of the bytes are carried by less than 5%

• f the largest flows [5,6].

 The high variability of Internet traffic negatively affect the end user experience in general due to the nature of TCP transport protocol and FIFO scheduling mechanism used in most network routers.

 small files data transfers often don’t reach to the steady state and congestion avoidance phase and their transfer terminate while they are in their slow start phase.  They are also prone to packet loss because of their small congestion window (mostly results in RTO).  This makes small file data transfers (short-lived flows) conservative in contrast to the aggressive long-lived flows.  In WLAN the situation is even worse due to the burstiness of bit errors in wireless channel.  short-lived flows suffer from higher variability of transfer time compare to the long-lived flows.

 LAS is a preemptive scheduling policy that favors short jobs without prior knowledge of the job sizes.  the mean response time of LAS highly depends on the job size distribution.  Based on LAS definition in network flows, the next packet to be served is the one that belongs to the flow that has received the least amount of service so far.  LAS significantly reduces the mean transfer time and loss rate of short TCP flows as compared to DropTail first in-first out (FIFO) scheduler.

Slow start phase for short TCP flows under both LAS and FIFO with no congestion

Slow start phase for short TCP flows under both LAS and FIFO with congestion

 A small increase in mean transfer time of large flows happens but under moderate load values a large flow under LAS is not starved when competing with short TCP flows.  However, the performance of the long-lived flow under LAS deteriorates severely when competing at high load with short flows.  One drawback of LAS is that one newly arriving long-lived flow can block all other existing long-lived flows until the time it receive the same service they achieved. In this case, all other long-lived flows get starved which causes severe unfairness among long-lived flows.

 The main idea of Threshold-based least attained service (TLAS) scheduling is to give the newly arriving flow service priority up to a certain threshold (e.g. 50 packets). Once the threshold is reached, FIFO scheduling is employed on this flow.  Authors in [2] showed that TLAS can guarantee fairness for short-lived flows as well as long-lived flows.  In our research we show that TLAS in unable to provide direction- based fairness. Because it gives the equal service to all flows until a certain threshold which normally is a small value and it behaves the same as FIFO for the rest of transmission period resulting in direction- based unfairness for most part of the transmission period (specifically for long-lived downstream flows).

 Barrows its main idea from TLAS and Selective Packet Marking-ACK filtering (SPM) [7,8].  Sets a minimum threshold for service that should be guaranteed for every network flow and it behaves the same as LAS with all packets below that threshold.  Once service threshold is reached for a flow it behaves with flow as SPM-AF (specified in [2]) does.  TLAS-SAF inherits TLAS capability to provide size-based fairness while employing SPM-AF characteristics for larger flows to provide direction-based fairness as well.

 In addition to giving priority to short-flows, TLAS-SAF also gives priority to the data packets which their congestion window of their flows is smaller than four.  Data packet loss in these flows result in retransmission timeout (RTO) while loss of other data packets will trigger fast retransmission mechanism.  It also frees the buffer space from repetitive ACKs of a flow relying

• n cumulative nature of ACKs.

TLAS-SAF procedure is as follow: 1. Gives service priority to newly arriving packets of a flow until a certain threshold (e.g. 50 packets). 2. Once threshold is reached, FIFO scheduling will be imposed on this flow . 3. If an arriving packet of a flow below threshold encounters to a full buffer, a packet of a below threshold flow which has received the maximum service so far (similar to TLAS) will be found and

• discarded. However, If received service for such a packet flow was

less than received service of arriving packet flow, the arriving packet will be discarded instead. 4. TLAS-SAF gives priority to data packets which their congestion window is smaller than four for the flows have already achieved the minimum threshold service.

5. In full buffer situation, if any new data packet with corresponding congestion window smaller than four arrives and its flow has already achieved the threshold service, it will be admitted in FIFO order at the queue but an ACK packet from above threshold flows will be discarded from the queue. 6. In addition, TLAS-SAF treats ACKs in such a way that at any given moment only one ACK from a certain flow which has already received the minimum threshold service exists in queue providing more buffer space for data packets. 7. For these flows when an ACK arrives to a full buffer, TLAS-SAF checks whether any other ACK from its flow exists in queue and if found it will simply get discarded.

TLAS-SAF flowchart

 Network Simulator (ns-2) is used for evaluation and performance analysis.  A new class is developed for each of queue management policies (LAS,TLAS,TLAS-SAF).

Queue class hierarchy LAS variant classes inherited from Queue class

Network topology used in ns-2 simulations

Common simulation setup parameters for all simulations.

Examined protocols Simulation area Propagation model Traffic type MAC layer Antenna Node to AP distance TLAS-SAF,TLAS,LAS,DropTail 670X670 m Two-ray ground TCP 802.11 Omni 5 m AP to Server link (common) Bandwidth Wired delay Queue policy 100 Mbps 2ms DropTail TCP setting (common) TCP protocol Packet size Segment per ACK Application TCP NewReno 1460 bytes 1 (ns-2 default) FTP

802.11 MAC setting (Common Parameters) MAC version Slot time SIFS Preamble length Short retry limit retransmissions Long retry limit retransmissions PLCP Header Length MAC basic rate MAC data rate Minimum Contention Window (CWMin) Maximum Contention Window (CWMax) 802.11b 20µs 10µs 72 bits 7 4 48 bits 1 Mbps 11 Mbps 31 1023

 𝑜 number of wireless nodes is defined which is varied from 6 to 46 nodes.  Half of the wireless stations transfer large files with a fixed file size of 1000 data packets (packet size equal to 1460 bytes) using FTP

• connections. Each of these stations downloads 10 such large files

sequentially.  The other half of the wireless stations generate small 10 packets file size download traffic with exponential on-off distribution of inter- arrival times with bust time of 500 ms and idle time of 500 ms as well with the same packet size as the large files. The small file transfers are repeated until all large file transfers are finished.

Simulation Parameters for size-based fairness evaluation.

Simulation time Number of nodes Default AP buffer size Nodes generating short-lived flows Nodes generating long-lived flows Number of large flows per station Number of short flows per station Large flow size Short flow size Large flow arrival time Short flow distribution Short flow burst time Short flow idle time End of large files transmission 6-46 nodes 50 packets Half of total Half of total 10 1 (repeating) 1000 packets 10 packets Sequentially Exponential ON-OFF 500 ms 500 ms

 Ten mobile stations are placed within 5 m of the AP.  Half of mobile stations are TCP senders. They upload data to the server using fixed-size data packets (1460 bytes).  The other wireless stations are TCP clients, downloading data from server using the same TCP data packet size as the upstream flows.  We run each simulation for duration of 300 seconds simulation time to determine steady state throughput. Since AP buffer size is a dominant factor determining the degree of direction-based fairness in WLANs we repeated our simulation under different AP buffer sizes starting from 5 packets to 200 packets.

Simulation Parameters for direction-based fairness evaluation.

Simulation time Number of nodes AP buffer size Nodes generating upstream flows Nodes generating downstream flows Number of flows per station 300 seconds 10 nodes 5-200 packets Half of total Half of total 1 (continuously)

1. Mean transfer time. (size-based evaluation) 2. Coefficient of Variation of Transfer Time . (size-based evaluation) 3. Round-trip Delay Time. (size-based evaluation) 4. Jain’s Fairness Index [9] . (direction-based evaluation)

Mean transfer time of a group of flows (e.g. short-lived flows) refers to the standard average of transfer time of all flows belonging to that specific group for the whole duration of a simulation period. It is calculated by where 𝑡et is a group of flows (e.g. short-lived or long-lived flows) which has 𝑜 number of flows and 𝑈( ) is the transfer time of flow .

In probability theory and statistics, the coefficient of variation (CoV) is a normalized measure of dispersion of a probability distribution. It is defined as the ratio of the standard deviation 𝜏 to the mean 𝜈. where 𝜏 is calculated by

Jain’s fairness index is calculated by where there are 𝑜 number of flows in system and refers to the throughput of 𝑗th flow.

CoV of transfer time for under DropTail/FIFO: (a) small files. (b) large files. (b) (a)

CoV of transfer time for under LAS, TLAS and DropTail/FIFO: (a) small files. (b) large files. (b) (a)

CoV of transfer time under proposed TLAS-SAF and other policies: (a) small files. (b) large files. (b) (a)

Mean transfer time under TLAS-SAF and other queue policies : (a) small files. (b) large files. (b) (a)

Average RTT of short-lived and long-lived flows : (a) DropTail/FIFO. (b) LAS. (c) TLAS. (d) TLAS-SAF

(b) (a) (c) (d)

 Similar to TLAS, TLAS-SAF is able to achieve a certain level of CoV

• f transfer time for small files (0.6 to 1.3) while not deteriorating

large file transfers. The reason why we choose TLAS-SAF over TLAS is due to the fact that TLAS is unable to guarantee direction- based fairness. (we will show it later on)  In terms of mean transfer time and Roundtrip time (RTT), TLAS-SAF performed as good as TLAS.

Jain’s fairness index of network flows under DropTail/FIFO policy at AP. 𝐶≥(𝜖MW+𝑂W)=5(20)+5(20)=200

Jain’s fairness index of network flows under LAS,TLAS and DropTail/FIFO policies at AP.

Jain’s fairness index of network flows under TLAS-SAF and other policies at AP.

In contrast to TLAS, our proposed TLAS-SAF queue management policy was able to provide fair bandwidth allocation among downstream and upstream flows for AP buffer sizes above 40 packets. It achieved a fairness index in range of between 0.93 and 1.0 for these AP buffer sizes outperforming LAS in term of overall fairness.

 We showed by simulations that TLAS-SAF is able to provide better service for short-lived flows while not deteriorating long-lived flows. These achievements in service improvement are in terms of mean and variation of transfer time as well as queuing delay experienced.  In addition, we observed based on simulation results that TLAS-SAF was able to provide a good degree of fairness among upstream and downstream flows achieving the closest fairness index to one for most

• f the AP buffer sizes compare to other studied queue management

policies.  We conclude that TLAS-SAF can be taken into account as a unique solution for both size-based and direction-based fairness issues in IEEE 802.11 WLANs while other proposed solutions in literature can

• nly perform well in one of these topics.

 More research activity should be taken to focus on fairness issues in high speed 802.11n WLANs and QoS oriented 802.11e WLANs.  adaptive dynamic buffer sizing can be seen as an interesting solution to prevent buffer spaces to become bigger in high speed WLAN.  we will try to focus on interaction of newer 802.11 MAC protocol and TCP variants to evaluate the performance of our proposed model.  In addition to simulations, test-bed experiment results must be taken into consideration to prove the performance of TLAS-SAF in real networks

[1] IEEE Computer Society LAN MAN Standards Committee, IEEE Standard for Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications. IEEE Std 802.11-1997 (The Institute of Electrical and Electronics Engineers, 1997). [2]

• Q. Wu, M. Gong, and C. Williamson, TCP fairness issues in IEEE 802.11 wireless LANs, Computer Communication

Journal, vol. 31, no. 10, pp. 2150-2161, 2008. [3]

• M. Bottigliengo, C. Casetti, C. Chiasserini, M. Meo, Short-term fairness for TCP flows in 802.11b WLANs,

Proceedings of IEEE INFOCOM, Hong Kong, China, 2004, pp. 1383–1392. [4]

• S. Pilosof, R. Ramjee, D. Raz, Y. Shavitt, P. Sinha, Understanding TCP fairness over wireless LAN, Proceedings of

IEEE INFOCOM, San Francisco, CA, 2003, pp. 863–872. [5]

• V. Paxson and S. Floyd, Wide-Area Traffic: The Failure of Poisson Modelling, IEEE/ACM Transactions on

Networking, 3:226–244, June 1995. [6]

• K. Claffy, G. Miller, and K. Thompson, The Nature of the Beast: Recent Traffic Measurements from an Internet

Backbone, In Proceedings of INET, July 1998. [7]

• C. Williamson, Q. Wu, A case for context-aware TCP/IP, ACM Performance Evaluation Review 29 (4) (2002) 11–23.

[8]

• Q. Wu, C. Williamson, A DiffServ framework for context-aware TCP/IP, submitted for publication, 2006.

[9]

• R. Jain, D. Chio, and W. Hawe. A Quantitative Measure of Fairness and Discrimination for Resource Allocation in

Shared Computer Systems, DEC RR TR-30I, 1984.