Determination of performance enhancement in IEEE 802.11 multi-rate WLAN.
The IEEE 802.11  or Wi-Fi is the specification defined by Institute of Electrical and Electronics Engineers (IEEE) for wireless local area networks (WLANs). The standard defines two layers; physical layer (PHY), which specifies the modulation scheme used and the signaling characteristics for the transmission through radio frequencies. The second layer is medium access control (MAC) layer. This layer determines how the medium is used . Different number of IEEE 802.11 standard exists (802.11, 802.11a, 802.11b, 802.11g, etc.).
IEEE 802.11 provides a way for allotting a part of the wireless channel bandwidth to some nodes using the Optional Point Coordination Function (PCF) or Hybrid Coordination Function (HCF) but the mandatory access protocol, Distributed Coordination Function (DCF), uses the Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) mechanism to share the channel in a fair way. It has been observed, however, that in some scenarios in a wireless setting, the method (DCF) results in significant performance degradation. Some nodes, in WLANs, may be far away from their access point therefore the quality of their transmission, due to signal fading, is low. In this situation, IEEE 802.11 products degrade the bit rate from the nominal highest rate to lower or even lowest nominal rate . That is, when a node detects a repeated transmission failure it reduces its bit rate. When there is at least a node with lower rate, IEEE 802.11 cell presents performance unfairness; the throughput of all nodes transmitting at higher rate is degraded by that slow rate host . The CSMA/CA is the basis for this anomaly as it guarantees that the probability of a long term channel access is equal for all nodes. Wireless devices adapt, normally, their transmission rate to the condition of the radio frequency (RF) channel. A device transmitting at a lower rate holds onto the RF channel longer than those with higher rates. This slows down the network and lowers the overall performance of the IEEE 802.11 WLANs. The paper  analyzed this anomaly of IEEE 802.11b and derived an important expression for the useful throughput but did not consider the fairness of the system. A time-based regulator algorithm that resides in the AP to provide time-based fairness by regulating packets is proposed in . The limitation of the algorithm is that it only schedules transmission of packet from or to a device if it has not used up all of its available time. A maximum transfer unit (MTU) adjustment proportionate with the transmission rate is proposed in . Their solution overcome the performance anomaly but did not address fairness problem in a multi-rate case. Authors in  proposed that the contention window (CW) be proportionate with the transmission rate on the expectation that the CWmin will be a configurable parameter in IEEE 802.11e extension. The useful throughput derived by  is utilized by , where a transmission rate-based packet size adjustment (TRPSA) scheme is proposed. Their scheme improves the performance anomaly in IEEE 802.11a/b/g by increasing the throughput and the fairness of the system. However, the scheme is not tested against IEEE 802.11n which is the extension of the previous standards. The aims of this work are to derive the throughput equation of both transmission time-based scheme (TTBS) and IEEE 802.11 WLANs, use it to calculate the theoretical maximum value of their fairness indices and validate same through simulation using Optimized Network Engineering Tool (OPNET) modeler. The objective is to compare the performance metrics of TTBS and IEEE 802.11 WLANs in a multi-rate situation.
2 PERFORMANCE ANALYSIS
In this section we derived the equation of one of the performance metrics of IEEE 802.11 WLANs, i.e., the throughput. We then used the equation to calculate the theoretical maximum value of Jain's fairness index (Jain et al. 1984; 1999); the other performance metric of IEEE 802.11 WLANs. Thereafter, we introduce our proposal called transmission time-based scheme (TTBS) to enhance and address the fairness issue in IEEE 802.11 WLANs.
According to  the time Ti for a single host or device to transmit a frame if propagation delay is neglected is given by:
[T.sub.i] = [t.sub.i] + [t.sub.ov] (1)
[t.sub.ov], [t.sub.i] are the overhead time and the frame transmission time respectively given by:
[t.sub.ov] = DIFS + [t.sub.pr] + SIFS + [t.sub.pr] + [t.sub.ack] and
[t.sub.i] = [S.sub.i]/[r.sub.i] (2)
[S.sub.i] is the frame size while [r.sub.i] is the transmission rate,
DIFS and SIFS are the distributed and short inter-frame space time given by:
DIFS = SIFS + 2 x aSlotTime (3)
The value of SIFS and aSlotTime is shown in table 1 for different PHYs and [t.sub.pr] is the physical layer convergence protocol (PLCP) preamble and header transmission time. [S.sub.i] is the frame size and [r.sub.i] the transmission rate.
When there is more than one host attempting to use or capture the wireless medium, the hosts have to contend for the medium. Therefore the overall transmission time has to account for the time spent in contention tcont by the hosts. This implies that equation (1) has to be modified. The equation can now be written as:
[T.sub.i] = [t.sub.i] + [t.sub.ov] + [t.sub.cont] (4)
Assuming that N hosts alternate transmission with all possible collisions and assuming further that multiple collisions are negligible , then, the total time during which all hosts transmit once is given by:
[T.sub.tot] = [[summation].sup.N.sub.j=1] + [P.sub.c] (N) x [t.sub.jam] x N (5)
[[summation].sup.N.sub.j=1] [T.sub.rj] = channel occupation time for all host with transmission rate [r.sub.j],
[P.sub.c](N) = Probability of collision,
[t.sub.jam] = average delay when packets form two node collision.
Therefore the channel utilization [U.sub.i] by any node is given by:
[U.sub.i] = [T.sub.i]/[T.sub.tot] (6)
According to  the useful throughput [P.sub.i] achievable by any host depends on the number of all hosts in the network and is given by:
[P.sub.i] = [t.sub.i]/[T.sub.i] (7)
Therefore, the throughput [X.sub.i] at MAC layer for any host with transmission rate [r.sub.i] is given by:
[X.sub.i] = [U.sub.i] x [P.sub.i] x [r.sub.i] (8)
Putting equations (6) and (7) in (8) yields:
[X.sub.i] = [T.sub.i]/[T.sub.tot] x [t.sub.i]/[T.sub.i] x [r.sub.i] (9)
but [t.sub.i] = [S.sub.i]/[r.sub.i] therefore equation (9) becomes:
[X.sub.i] = [S.sub.i]/[T.sub.tot] (10)
In IEEE 802.11 all hosts transmit equal-sized frames, therefore S, is the same for all hosts. This causes the performance anomaly in a multi-rate multi-node situation as all nodes get the same throughput regardless of their transmission rates.
We now introduce the Jain 's fairness index [9-10] which measures fair allocation of resources (in our case the wireless medium) in a distributed system. The fairness index (FI) is bounded between 0 and 1. If the index approaches 1 the system is very fair otherwise it is not. The FI in terms of transmission rate r, and throughput X, is given by:
FI = [([[summation].sup.N.sub.i=1] [r.sub.i]/[X.sub.i]).sup.2]/N [[summation].sup.N.sub.i=1][([r.sub.i]/[X.sub.i]).sup.2] (11)
We have shown in (10) that [X.sub.i] is the same for all hosts in IEEE 802.11 WLANs. Therefore equation (11) becomes;
FI = [([[summation].sup.N.sub.i=1] [r.sub.i]).sup.2]/N [[summation].sup.N.sub.i=1][([r.sub.i]).sup.2] (12)
Let us take the example of IEEE 802.11b which specifies four different data rates of 1, 2, 5.5 and 11Mbps to calculate the value of the fairness index. The number of rates is four therefore N = 4. This gives us;
FI = 0.6084
This implies that the maximum theoretical value of FI achievable by IEEE 802.11b wireless system is 0.6084. That is the system is around 60% fair.
2.1 Transmission Time-Based Scheme (TTBS)
We now propose that instead of all hosts to have equal-sized frames, let the hosts have equal transmission time. This, we believe, will improve the network performance from time-based fairness point of view. That is, let [t.sub.i] be the same for all host. But [t.sub.i] is given by:
[t.sub.i] = [S.sub.i]/[r.sub.i] (13)
Let the frame size and transmission rate of the fastest node be [s.sub.max] and [r.sub.max] respectively. Further, let [s.sub.i] and [r.sub.i] be the frame size and transmission rate of any other node in the network. Since transmission time is the same for all hosts in our proposal, therefore equation (13) can now be written as:
[t.sub.i] = [s.sub.i]/[r.sub.i] = [s.sub.max]/[r.sub.max]
This implies that;
[s.sub.i] = [s.sub.max]/[r.sub.max] x [r.sub.i] (14)
Equation (14) shows that if transmission time is equal then the frame size has to change. In other words for any given host its frame size is proportional to its transmission rate. Therefore according to equation (14), equation (10) is no longer constant. i.e., throughput is not the same for all hosts in our scheme and given smax, rmax and [r.sub.i] the frame size of any other node [s.sub.i] can be calculated.
Now invoking the fairness index by substituting equation (10) in (11) yields:
FI = [([[summation].sup.N.sub.i=1] [r.sub.i]/[S.sub.i]).sup.2]/N [[summation].sup.N.sub.i=1][([r.sub.i]/[S.sub.i]).sup.2]
Now putting equation (13) in (15) gives:
FI = [([[summation].sup.N.sub.i=1] 1/[X.sub.i]).sup.2]/N [[summation].sup.N.sub.i=1][(1/[X.sub.i]).sup.2]
In our scheme [t.sub.i] is the same for all host, therefore equation (16) yields:
FI = [(4/t).sup.2]/4X4 [(1/t).sup.2] = 1
[therefore] FI = 1 (17)
Equation (17) shows that the IEEE 802.11b wireless system can be absolutely fair using our scheme.
We have designed and simulated twenty two (22) set of scenarios with simulation run-time of hundred seconds (100s) for each. The nodes communicate with an access point (AP) and transmit a constant bit rate (CBR) traffic to the AP. IEEE 802.11 standard specifies that all node transmit an equal-sized frame so we adopt a frame size of 1000byte. We further adopt IEEE 802.11b version of the standard for the scenarios as it specifies only four (4) different transmission rates of 1, 2, 5.5 and 11Mbps unlike IEEE 802.11a/g which specify up to eight (8) different transmission rates and only basic access method was used for the simulation. Optimized network engineering tool (OPNET) modeler academic edition version 17.5 was used for the simulations. In designing the scenarios in the modeler, packet inter-arrival time was set according to the transmission time of each node. Packets on-off generation time were set to 1 and 0 respectively while start time was set to 0.1. All other values used were the modeler's default unless stated otherwise.
In the fi rst four sc enarios nodes A, B, C and D, as shown in figure 1, transmit with the same transmission rate of 1, 2, 5.5 and 11Mbps. In the fifth scenario each node transmits with a different transmission rate. That is 1, 2, 5.5 and 11Mbps respectively. Designing and simulating our proposal, TTBS, requires that each node has its frame size proportional to its transmission rate i.e., 91, 1 82, 5 00 and 1000byte for 1, 2, 5.5 and 11Mbps in that order. Equation (14) was used to arrive at the given frame size values.
Now using the frame size values above we simulated another set of four scenarios where we varied the nodes number according to the transmission rates. That is in the first scenario we varied the number of 1Mbps node from 2-14 and kept the number of 2, 5.5 and 11Mbps nodes constant. We repeated this with 2, 5.5 and 11Mbps nodes. These set of scenarios were repeated with IEEE 802.11b for comparison.
Next we checked the influence of packet size variation by fixing the packet sizes of three nodes and allowing the other node to take values form 100-1500byte. We now compared TTBS against IEEE 802.11a, g and n in a multi-rate condition. The respective packet sizes for 802.11a/g/n are given in table 2. Lastly we compared TTBS and 802.11 in a dynamic multi-rate situation, where the four nodes (A, B, C and D) move towards the access point with a speed of 1, 1, 2 and 2m/s respectively.
4 RESULTS AND DISCUSSIONS
In this section the results of the simulation scenarios described in section 3.1 are presented and discussed.
4.1 Performance Anomaly when Multi-rate Nodes Coexist
Figure 2 shows that in a single-rate situation the wireless system is very fair and the throughput of the individual node is almost equal. We adopt 11Mbps single-rate, as figure 2 (a) reveals that the highest throughput is attained in 11Mbps single-rate scenario, for comparison with multi-rate where nodes A, B, C and D transmit with the rate of 1, 2, 5.5 and 11Mbps respectively. The average throughputs of A, B, C and D were found to be 1.95, 2.00, 1.97 and 1.97Mbps in that order and the system throughput was 7.9Mbps in figure 2. Figure 3 (a) shows the performance anomaly clearly as the throughputs of nodes A, B, C and D drop to 0.54, 0.47, 0.51 and 0.57Mbps and system throughput downs to just 2.10Mbps. Although the transmission rate of node D does not change, its throughput drops to 0.57Mbps about 71% loss and system throughput drops by 73%. The fairness index is around 1, most fair, when only single-rate multi-node is present, immediately multiple-rate coexist the index drops sharply to about 0.60 as predicted by equation (12). This clearly revealed the nature of the fairness issue in IEEE 802.11 multi-rate WLANs.
Comparing TTBS with 802.11 under multi-rate condition shows that the system throughput raises from 2.10Mbps in 802.11 to 3.26Mbps in TTBS about 36% appreciation (figure 4a). The system fairness rises to 1 unlike in 802.11 where it hovers around 0.6. This result indicates that the system performance under TTBS improves.
4.2 Influence of Node Number Variation on the System Performance
In figure 5 (a) and (b) we notice the influence of node number variation with respect to 1Mbps node (the node with the least transmission rate) on the system. It is as expected; the nodes in TTBS scenario have different throughput while maintaining almost equal throughput in 802.11. Even though the fairness index of TTBS decreases a bit, the system is still very fair unlike in 802.11. The system throughput decreases in both cases as only the number of nodes with the least transmission rate increases. When considering node number variation, the throughput and fairness index of the node under investigation are average values for any given number of the said node.
Figure 6(a) and (b) show that the system aggregate for 802.11 increases while that of TTBS decreases because all the nodes have nearly equal throughput in 802.11, so if the number of any node other than the node with the least transmission rate increases the system throughput appreciates unlike in TTBS where nodes have different throughput therefore the system throughput only appreciates when the higher nodes numbers increase. Figures 6, 7 and 8 support the argument we just stated. They also highlight the fact that the wireless system is most fair when the number of the node under variation is just two (2) beyond that the fairness index begins to fluctuate downward a bit as the wireless system allocates more resources to nodes with same transmission rate.
4.3 Influence of Packet Size Variation on System Performance
Now we study the influence of packet size variation on the network. Figure 9a reveals that whenever 1Mbps node has equal-sized packet as any other node it achieves the same throughput as that particular node. We further see that the throughput of the 1Mbps node appreciates as its packet size increases while the throughput of other nodes decreases. This is due to the fact that 1Mbps being the slowest of all captures the channel longer than the other nodes. The fairness index (b part of same figure) and system throughput are highest, 0.97 and 3.25Mbps, when the nodes packet sizes-transmission rates ratio is 91: 182: 500: 1000 = 1: 2: 5.5: 11. This confirms that TTBS can effectively improve the network performance.
We expect, in figure 10a, the 2Mbps to have same throughput as 1, 5.5 and 11Mbps nodes when its packet sizes are 91, 500 and 1000byte, our expectation was only met at 91byte. Still in the same figure, 2Mbps node gets same throughput with 5.5 and 11Mbps nodes at 600 and 1200bytes because 2Mbps node is twice faster than 1Mbps node therefore it cannot achieve same throughput with 5.5 and 11Mbps nodes at the same point the 1Mbps node did, it has to slow down to the speed of 1Mbps node by adjusting its packet size. The 2Mbps node coincides with 5.5Mbps node at 600byte but 11Mbps is twice faster than 5.5Mbps node, hence the coincidence of 2 and 11Mbps nodes at 1200byte.
Even though, figure 11a, the throughput of 5.5Mbps node drops when it assumes the packet sizes of lower rate nodes its throughput is still above theirs; it is 5.5 and 2.75 faster than 1 and 2Mbps nodes respectively. When the 5.5Mbps node assumes the packet size of 11Mbps node (1000byte) its throughput is less than that of 11Mbps node; it's twice slower than 11Mbps node. Since 11Mbps node is faster than 1, 2 and 5.5Mbps nodes, its throughput should be higher when it assumes their packet sizes. This is illustrated in figure 12a.
4.4 Performance Comparison of TTBS with IEEE 802.11a/g/n
When the throughputs of IEEE 802.11a and TTBS were compared in figure 13a their average values were found to be 14.25 and 18.08Mbps respectively. The TTBS outperforms 802.11a by about 21%. Their fairness indices stand at 0.96 and 0.68, a difference of 29% in favor of TTBS.
TTBS still improves the network performance when compared with 802.11g (figure 14). The average values of their throughputs are 15.60 and 13.19Mbps while their fairness indices stand at 0.94 and 0.70 respectively. TTBS increases the system throughput and fairness by 15 and 26% in that order thereby improving the network performance considerably.
Figure 15 reveals somehow a slightly different picture as both TTBS and 802.11n achieve similar throughput of 16.59 and 16.63Mbps respectively. Actually the throughput of 802.11n is higher by just 0.003%. This is expected as 802.11n is the extension of 802.11a/b/g; there has been a lot of improvements in 802.11n as it utilizes multiple-inputs multiple-outputs (MIMO) antennas where only the signal with maximum strength is selected by the receiver. As expected TTBS achieves 0.64 fairness index increasing the system fairness by 63%.
4.5 Performance Comparison between TTBS and 802.11b under Multi-rate Dynamic Condition
Figure 16 compared TTBS and 802.11b in a dynamic multi-rate condition. TTBS records a throughput of 2.62Mbps while 802.11b achieves just 2.09Mbps. The TTBS method improves the system throughput by 20%. In terms of system fairness TTBS scores a fairness index of about 0.97 while 802.11b records just 0.62, 36% less than TTBS.
We have calculated the theoretical maximum values of IEEE 802.11b and TTBS fairness indices. The values of the indices were found to be 0.61 and 1 respectively. This result shows that IEEE 802.11b wireless system is just about sixty one percent (0.61%) fair in sharing the wireless medium among the participating nodes whereas TTBS, if adopted, makes wireless system hundred percent (100%) fair.
We then compared the performance metrics (throughput and fairness index) of transmission time-based scheme (TTBS) and IEEE 802.11 WLANs. We observed that, the percentage increase in throughput and fairness index of TTBS were found to be 21% & 29% when compared with IEEE 802.11a, 35% & 39% in comparison with IEEE 802.11b and 15% & 26% when compared with IEEE 802.11g respectively. TTBS throughput percentage was lower by 0.003% in comparison with 802.11n but its fairness index is higher by 63%. In a dynamic multi-rate scenario TTBS throughput and fairness index percentages were also found to be higher by 20% and 36% in that order when compared with IEEE 802.11b.
The results obtained proved that the new scheme (TTBS) can actually enhance the network performance effectively.
This paper being part of an M.Sc research work is sponsored by Federal University, Kashere (FUK) through TETFund study fellowship.
[1.] IEEE 802.11 Standard: Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications. ANSI/IEEE 802.11 (1999).
[2.] Gupta, N., Kumar, P.R.: A Performance Analysis of the 802.11 Wireless LAN Medium Access Control. Communications in Information and systems, vol. 3, No. 4 (2004), pp. 279-304.
[3.] Holt, A., Huang, C.: 802.11 Wireless Networks: Security and Analysis. In A.J. Samees (ed.), Computer Communications and Networks. Cranfield University, Swindon, UK: Springer (2010).
[4.] Heusse, M., Rousseau, F., Berger-Sabbatel, G., Duda, A.: Performance Anomaly of 802.11b. Proceedings of IEEE INFOCOM, (2003), pp 836-843.
[5.] Tan, G., Guttag, J.: Time-based Fairness Improves Performance in Multi-rate WLANs. USENIX Annual Technical Conference, General Track (2004), pp. 269-282.
[6.] Yoo, S., Choi, J., Hwang, J., Chuck, Y.: Eliminating the Performance Anomaly of 802.11b. In P. Lorenz and P. Dini (Eds.): ICN 2005, LNCS 3421. Springer-Verlag Berlin Heidelberg, pp. 1055-1062.
[7.] Kim, H., Yun, S., Kang, I., Bahk, S.: Resolving 802.11 Performance Anomalies through QoS Differentiation. IEEE Communications Letters, vol. 9 (2005), No. 6, pp. 655-657.
[8.] Hu, S., Li, J.: Pan, G. (2014). Performance and Fairness Enhancement in IEEE 802.11 WLAN. International Journal of Electronics and Communications, vol. 68 (2014), pp. 667-675.
[9.] Jain, R. K, Chiu, D, W., Hawe, W. R.: A Quantitative Measure of Fairness and Discrimination for Resources Allocation System. Submitted for Publication to ACM Transaction on Computer System, (1984) Dec-TR-301.
[10.] Jain, R., Durresi, A., Babic, G.: Throughput fairness index: An Explanation. ATM Forum Contribution, No. 99-0045 (1999), pp. 1-13.
Ibrahim Abubakar Alhaji
Department of Physics, Federal University, Kashere
P.M.B. 0182, Gombe, Nigeria
Bello Idrith Tijjani, Mutari Hajara Ali
Department of Physics, Bayero University, Kano
P.M.B. 3011, Kano, Nigeria
Table 1: Slot Time and SIFS Values for Each PHY PHY aSlotTime (s) SIFS (s) FHSS 50 28 DSSS 20 10 OFDM 9 16 Table 2: 802.11a/g/n Rates with Their Proportional Packet Sizes S/N 802.11a/g 802.11a/g 802.11n rate (Mbps) packet size rate (Mbps) proportiona l to rate (byte) 1 6 111 6.5(base)~60(max) 2 9 167 13(base) ~ 120(max) 3 12 222 19.5(base) ~ 180(max) 4 18 333 26(base)~240(max) 5 24 444 39(base)~360(max) 6 36 667 52(base)~480(max) 7 48 889 58.5(base)~540(max) 8 54 1000 65(base)~600(max) S/N 802.11n packet size proportional to rate (byte) 1 100 2 200 3 300 4 400 5 600 6 800 7 900 8 1000
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||wireless local area network|
|Author:||Alhaji, Ibrahim Abubakar; Tijjani, Bello Idrith; Ali, Mutari Hajara|
|Publication:||International Journal of Digital Information and Wireless Communications|
|Date:||Oct 1, 2015|
|Previous Article:||A framework for statistical characterization of indoor data traffic for efficient dynamic spectrum access in the 2.4 GHz ISM band.|
|Next Article:||Elucidating digital information and wireless communications technology for an electronic working logs and applications.|