Coexistence Mechanism Using Dynamic Fragmentation for Interference Mitigation between Wi-Fi and Bluetooth Alex Chia-Chun Hsu

David S. L. Wei

C.-C. Jay Kuo

Dept. of Electrical Engineering University of Southern California Los Angeles, CA 90089-2564 [email protected]

Dept. of Computer and Information Sciences Fordham University Bronx, NY 10458 [email protected]

Dept. of Electrical Engineering University of Southern California Los Angeles, CA 90089-2564 [email protected]

Abstract— In this paper, we present a non-collaborative coexistence mechanism using dynamic fragmentation that functions very well at the Wi-Fi in the presence of Bluetooth interference. The proposed scheme tries to optimize MAC layer packet length such that the Wi-Fi device has better chance to avoid the interference caused by Bluetooth devices. The mechanism dynamically adjusts the fragmentation level based on the current packet error rate (PER). We developed an analytical model that provides the MAC the necessary information (in terms of PER) to decide the right time for further packet fragmentation. The developed model is also employed to measure the throughput of the Wi-Fi device. Simulations are also performed to validate the developed model. We show that our coexistence mechanism could significantly improve the performance of Wi-Fi in both throughput and transmission delay, though there is only slight performance improvement at the Bluetooth side.

I. I NTRODUCTION The ISM band will soon be populated by various kinds of wireless devices. Most of these devices are likely to be in the area of wireless local area networking (WLAN) using the Wi-Fi technology [1] and wireless personal area networking (WPAN) using the Bluetooth technology (BT) [2]. Since WLAN and WPAN are complementary rather than competing technologies, it is quite likely that Wi-Fi and Bluetooth devices will appear in the same place. Since both devices use the same frequency band, there will be high potential of interference [3] [4]. To realize the ubiquitous communications environment through shared frequency band, devices need to have not only an efficient mechanism to access the unlicensed frequency band, but also a proactive mechanism to detect and mitigate the interference caused by other types of coexisted wireless devices. Recognizing the importance of coexistence and the need for studying the impact of interference on the throughput performance of those devices, IEEE has created a coexistence task group called IEEE P802.15.2 [5]. The IEEE 802.15 TG2 has defined two classes of coexistence mechanism, namely collaborative solution and non-collaborative one. Though the classification is mainly for the mechanisms developed for BT,

the principle is also valid for other coexisting scenarios. Collaborative solutions work only when coexisting networks are allowed to exchange information. Examples for collaborative solutions for coexistence of Wi-Fi and BT are META (MAC Enhanced Temporal Algorithm) or TDMA (Time Division Multiple Access) schemes [5]. Collaborative solutions are feasible only when the two systems are collocated on the same device and retrained by a centralized controller, and has to be under the assumption that there is no interference from non-collocated devices. On the other hand, non-collaborative solutions are the only choice when there is no way to exchange information between coexisting networks, which is likely to be the case in practical situation. With non-collaborative coexistence mechanism, each device simply takes its own maneuver to reduce the interference. With the rapid frequency hopping mechanism and broader operating frequency band, BT is more capable of avoiding interference caused by the Wi-Fi devices. Nevertheless, there have been some works of non-collaborative solutions that were developed to further strengthen the interference mitigation capability of BT. These works include AFH (Adaptive Frequency Hopping), BIAS (Bluetooth Interference Aware Scheduling) [6] [7], and D-OLA (Data-OverLap Avoidance) [8]. Basically, these mechanisms adopt the scheme of transmission scheduling to avoid overlapping in frequency. The approach is to distinguish the good channels from the bad ones and then keep the hopping sequence at the “good” channels more frequently than at the “bad” ones. On the other hand, Wi-Fi is more vulnerable to the interference caused by the BT devices due to its longer data packet and lack of agility to frequency changing. Thus it is desperately needed to develop coexistence mechanism on the Wi-Fi side. A scheduling mechanism, V-OLA (VoiceOverLap Avoidance) [8], has been proposed to avoid the interference from the BT voice traffic by squeezing Wi-Fi transmission into the idle time between consecutive BT voice packets. However, it works only when BT has voice traffic only. Otherwise it requires coexisting BT devices to run D-

OLA [8]. Therefore, it’s not a pure non-collaborative solution. Meanwhile, packet fragmentation has been proposed as an effective way to mitigate interference. However, we found that what proposed in [9] requires too much complexity to obtain the optimal solution and gradually change the packet length may not gain the immediate performance boost we need. We will show that determining the right timing to do fragmentation is more practical and effective than determining the optimal fragmentation. In this paper, we focus on the non-collaborative solution on the Wi-Fi side. Our proposed mechanism, called Dynamic Fragmentation Algorithm (DFA), dynamically adjust the fragmentation level based on the current level of packet error rate (PER). Thus the mechanism is triggered only when needed. The version for the scenario with mobility is called DF Am , and the version for the scenario of static network is called DF As . We first develop an analytical model to evaluate the PER of Wi-Fi and BT devices under interference, and then show that the model can be employed to effectively determine the right time for further fragmentation or move back to the previous state of unfragmentation. Furthermore, with DFA, 802.11 MAC is able to distinguish whether the failed transmission is an interference caused by a BT device or a collision caused by a Wi-Fi station, and thus can take different procedures so that the higher throughput can be achieved. Simulation results show that our DFA could help reduce interference between Wi-Fi and BT, and thereby significantly improve the performance of Wi-Fi in both throughput and transmission delay, though there is only slight performance improvement at the Bluetooth side. The rest of the paper is organized as follows. Section II presents the analytical model used to estimate the PER of the Wi-Fi and BT. Section III further develops the model for the estimated transmission time for Wi-Fi with variable fragment level and fragment length, and then combining with the PER model, we could formulate the cost equations of all kinds of fragmentation. Section IV first describes the simulation setup and system parameters, and then presents the simulation results to validate our analytical model and to show the performance improvements on the throughput and transmission delay. The conclusions are drawn in Section V. II. A NALYTICAL M ODEL FOR I NTERFERENCE Interference between Wi-Fi and BT occurs when the transmissions of the two systems overlap both in frequency and in time. Whenever the signal strength of either is over the SNR threshold of the other, it results in a packet loss. Depending on the separation distance, transmission power/carrier sensing, and channel conditions, three scenarios of packet loss could happen: either a BT packet loss, or a Wi-Fi packet loss, or both. Though this is also a kind of collision between data packets, however, the nature is very different from the collision between Wi-Fi stations. Therefore, throughout the rest of this paper, we will refer the packet loss caused by the interference between BT and Wi-Fi as an interference incident and reserve

Fig. 1.

Wi-Fi and BT packets.

the term collision for the packet loss incident caused by Wi-Fi stations alone. Model of Packet Error Rate We start establishing our analytical model of interference with analysis on packet length of BT and Wi-Fi. Normally, a Wi-Fi packet is several BT time slots long. Let TW be the time duration of Wi-Fi packet transmission1 , TB be the BT time slot and TBA be the active time within each BT time slot. Let t be the time interval between the beginning of WiFi packet and the beginning of the first overlapped BT time slot (See Fig. 1 as an example). Then we could calculate the number of BT time slots that are overlapped with a Wi-Fi packet. Let the number be N 2 . ( e if t ≤ d TTW eTB − TW , d TTW B B N= (1) TW d TB e + 1 otherwise. Pf is the probability that a BT device hops on the frequencies that would interfere Wi-Fi transmission. There is an inactive period in each BT time slot, and we express the utilization of a BT time slot as σ . The probability of a BT packet in a time slot, which is also the activity level of BT, is denoted by τ BT , which represents the traffic load of WPAN. With single WPAN and single WLAN, we could express the packet error rate of a Wi-Fi station under interference as P ERW i−F i = 1 − (1 − Pf τ BT σ )N ≈ N Pf τ BT σ , where σ = TTBA B

(2)

Let τ W i−F i be the activity level of Wi-Fi, then the packet error rate of a BT device under interference could be expressed as P ERBT = Pf τ W i−F i

(3)

III. DYNAMIC F RAGMENTATION A LGORITHM (DFA) According to our observation, there are two fundamental problems that Wi-Fi has to cope with in order to mitigate interference in coexistence environment. First, a Wi-Fi station finds no way to tell if a packet loss is due to collision or interference. As a result, CSMA/CA would treat all packet loss the same way, double the backoff window and retransmit 1T W contains DATA and ACK, and if any portion of the duration is interfered, it will cause retransmission. 2 For variable packet length, T W means E[TW ] and N means E[N ].

the packet. This leads to the second problem, i.e. CSMA/CA is not efficient to deal with the interference problem. The collision avoidance mechanism is designed to solve the traffic congestion among Wi-Fi stations. With longer backoff time, the traffic is expected to average out over the time and thus lower the chance of collision. However, the chance of interference won’t be lowered by increasing the backoff time, and the CSMA/CA will simply introduce more unnecessary overhead but not decrease the chance of interference. Therefore we need to devise a new mechanism to solve the problem of packet loss due to interference, and this motivates us to develop DFA algorithm. Observe that the ratio of Wi-Fi packet length to the BT time slot is crucial to Wi-Fi packet error rate (PER). We thus want to develop an algorithm that can adjust the Wi-Fi packet length (using the existing fragmentation function of 802.11 [1]) based on the latest PER information to reduce the interference. In other words, with our DFA algorithm, the legacy 802.11 MAC will be empowered with the ability to cope with the problem of interference at run time. Be informed that our algorithm aims at reducing the interference rate but not the collision rate. The job of reducing collisions is till on the shoulder of the CSMA/CA. Though there are only two states in the current version of our algorithm, the algorithm can be extended to a version of more states. In State 1, DATA is transmitted in one piece without fragmentation and in State 2, DATA is divided into η fragments and are transmitted sequentially. The system collects the packet error rate (PER) information periodically in a fixed time interval. For state transition, we need to assign two thresholds, a move-up threshold P2 and a move-down threshold P1 for the two-state version. At the beginning of each time interval of PER collection, the station checks if the latest PER is higher than P2 . If so, moves one state up if a higher state is available. Otherwise, stay in the same state. Here, a state up means further fragment the packet. If the latest PER is lower than P1 , then move one state down if a lower state is available. Otherwise, stay in the same state. The next step is to determine the threshold P1 and P2 , which is a crucial step in the algorithm. Before doing so, we need to explain how fragmentation works. Fig. 2(a) shows the basic packet transmission of legacy 802.11 without fragmentation. It first waits for DIF S, then waits for the end of Backoff Window BW (or contention window), and then data is sent in one piece, and then an ACK is received. Fig. 2(b) shows the packet transmission with fragmentation. In this example, DATA is divided into two fragments (η = 2), namely DAT A 1 and DAT A 2. The first fragment is sent with full contention mechanism, then after SIF S, the second fragment is sent without contention. Fragments are ACKed separately. Fig. 2(c) shows when some fragment is failed to be received, it has to be retransmitted with contention mechanism. The bottom line is prior fragment has to be successfully received before any attempt of next fragment. For the following transmissions, the contention mechanism could be omitted if it’s not a retransmission. Like [8], to simplify the discussion, we assume

Fig. 2. Packet Fragmentation: scenario (a) legacy 802.11 with no fragmentation, scenario (b) two fragments without any retransmission, scenario (c) retransmission on the first fragment due to failed ACK 1

that there is no hidden node problem and thus did not take RTS/CTS into account, as we only consider the performance affected by interference. Nevertheless, according to our study, even if we take the RTS/CTS into account, it will not affect the outcome. We begin our derivation of P1 and P2 with time analysis of a successful transmission. For a successful transmission with n fragments, like scenario (b), the total transmission time with no retransmission Ts is TDAT A ) (4) n Here Toh = Th + 2 ∗ SIF S + TACK , which is the fixed overhead, and Th represents the time to transmit the header of a packet. TDAT A and TACK represent the actual transmission time of DAT A and ACK, which is the packet length divided by transmission rate. BW represents the backoff window. DIF S and SIF S represent the corresponding inter-framespace. Next, if there is any transmission failure, a retransmission would take place, like scenario (c). The time needed for a single retransmission Tr is Ts = DIF S − SIF S + BW + n(Toh +

TDAT A (5) n At last, we could express the total transmission time of a packet which is divided into n fragments and eventually suffers a total of R retransmissions. The total transmission time would be the successful transmission time of n fragments plus R times the single retransmission time, i.e. Tr = DIF S − SIF S + BW + Toh +

Ts + R × Tr = (R + 1)(DIF S − SIF S)+ X TDAT A ) BWi + (n + R)(Toh + n

(6)

1+R

Now we are ready to evaluate the thresholds. As a state transition algorithm, the timing of state transition is crucial. In DFA, the decision on state transition depends on whether the transition could lower the expected transmission time. When total transmission time is expected to be lowered, a state transition is necessary. Otherwise, stay in the same state. From

State 1 to State 2, the number of fragments changes from n to n0 (= ηn). Assume that the total retransmission counts would change from R to R0 , the condition of transition could be expressed as X

(R + 1)(DIF S − SIF S) +

A + Toh ) BWi + (n + R)( TDAT n

1+R X

>

A BWi + (n0 + R0 )( TDAT + Toh ) n0

(R0 + 1)(DIF S − SIF S) +

1+R0

Taking expected values, we have (E[R] − E[R0 ])(DIF S − SIF S) + (E[R] − E[R0 ] − (η − 1)n)Toh +(E[R] −

E[R0 ] TDAT A ) n η

+ E[

X

BWi ] − E[

X

BWi ] > 0

R0

R

(7) To calculate the threshold, we need to find all the expected values in the equation. We begin with E[R], where R is the n X total number of retransmission. We have E[R] = E[ Ri ], 1

where E[Ri ] represents the expected value of retransmission count for each fragment, and n times of the value would be the expected value of overall retransmission count. Assuming geometric PDF distribution of Ri , we have

Fig. 3.

Threshold from legacy 802.11 to DF Am,s

Since the backoff window only doubled up to the upper bound CWmax , there are two cases for analysis. Case I, E[Ri ] ≤ b − a X X E[ BW ] = n × E[BWi ] R

fRi (k) = pk (1 − p) where p = P ERW i−F i , k is the retransmission count

(8) Case II, E[Ri ] > b − a

Pf , τ BT , σ will remain Then E[R] = n × E[Ri ] = the same after state transition, and only n and N will change after fragmentation. With Eq. 2, we could predict p0 , which is the PER after state transition, and E[R0 ]. np 1−p .

P ERW iF i P ERW iF i0

E[R0 ] =

=

n0 p0 1−p0

p p0

=

=

E[Ri ]

= 12 (2a+1 (2E[Ri ] − 1) − E[Ri ]) × nTslot

N0 N p, then 0 ηnp p = κ−p = Nn −p N0 0 N n N0 , η = n

N N0

→ p0 =

0 n0 N N p N0 1− N p

where κ =

(9)

X Next, before we begin the derivation of E[ BWi ] and R X E[ BWi ], we need to define some parameters. Let i be the R0

retransmission counts for any single fragment, and a and b be constants in CWmin = 2a − 1 and CWmax = 2b − 1, where CWmin and CWmax are the minimum and maximum size of the contention window. Then the length of backoff window of the ith retransmission can be expressed as BWi ∈ [0, 1, 2, ..., 2i+a − 1] × Tslot , and E[BWi ] = 12 [2i+a − 1] × Tslot

(10)

Let E[BWi ] be the average backoff window of the ith X retransmission for a fragment. Then E[ BWi ] = n × R X E[BWi ], where E[Ri ] represents the average number E[Ri ]

of retransmission of a fragment and

X

E[BWi ] represents

E[Ri ]

the average total length of backoff window of any fragment.

X X E[ BW ] = n × E[BWi ] R

E[Ri ]

= 12 (2a+1 (2b−a − 1) − 2b (b − a) + (2b − 1)E[Ri ]) × nTslot Now all the terms in Eg. 7 are known, we could plug in all the expected values and calculate the threshold, i.e. P1 , P2 , using Eg. 7. We used the system parameters from Tables I and II to evaluate the theoretical threshold from State 1 (legacy 802.11 with no fragmentation) to State 2m,s (DF Am,s ). Figs. 3 is drawn from the analytical throughput under different PER levels. It’s clear that if the PER is higher than the crossing point, fragmentation will improve throughput, and below that point, fragmentation would have no positive effect. These thresholds are sensitive to the packet header, which accounts for majority of fragmentation cost, and the ratio of the number of overlapped BT time slots to Wi-Fi fragment of current and previous state, which is κ in Eq. 9. Normally during a Wi-Fi transmission, there should be no collision in the middle of a transmission if each node is static. The only collision scenario is when two Wi-Fi stations choose the same size of backoff window, and they thus will end up starting their transmission at the same time. In other words, if there is no mobility or hidden node problem, collision either happens from the beginning of the transmission or it doesn’t happen at all. For the case of interference, the picture is quite different. Since there is no time-wise collision avoidance mechanism on BT side, interference could happen at any time during a Wi-Fi

Fig. 4. scenario (c) DFA with mobility concern (DF Am ), scenario (d) DFA with static nodes (DF As )

Since collision may incur the retransmission of the first fragment, we need to keep the backoff window, r is the retransmission count for the first fragment, and R−r would be the retransmission count for the subsequent fragments. Then we follow the same steps to find P1 , P2 using Eg. 7. By removing the unnecessary backoff windows, we lower the price of fragmentation. Thus the threshold of entering the fragmentation state is lowered, see the difference in Figs. 3. IV. S IMULATIONS

transmission. And unlikely it will happen at the beginning of a transmission. Without DFA, the CSMA/CA can not perceive this fine feature. With DFA, we could observe the transmission status of each fragment, if the first fragment of a fragments array suffers failure, it could be a collision or a interference. On the other hand, if the subsequent fragments encounter failure, it is most likely due to interference. In the CSMA/CA with DFA, if a station enters a fragmentation state and observes no decline on the PER or no transmission failure on the second fragment or beyond, it would recognize that the high PER is caused by collisions only and switch back to nonfragmentation state. Since under that scenario, fragmentation will fail to reduce PER but just decreases the throughput by introducing unnecessary overhead. In order to improve throughput, one could either cut down overhead or abate retransmissions. For fragmentation, there is a tradeoff between retransmission and overhead. The further we do fragment, the more overhead will certainly get. However, the less retransmission is expected. When the gain from reducing the retransmission couldn’t balance off the loss from the increased overhead, fragmentation will do no good, and that’s why in previous subsections we carefully analyze the cost of fragmentation to guarantee the improvement. We already explained that except for the first fragment, the retransmissions of the subsequent fragments are most likely caused by interference, and the backoff windows assigned by collision avoidance mechanism for those subsequent fragments will just introduce unnecessary overhead. Since the probability of interference wouldn’t be decreased by introducing these backoff windows, we could simply bypass them. See Fig. 4 as an example. Now with the new procedure, when encounter a failed transmission, sender would retransmit the fragment immediately after ACK timeout without waiting for backoff window. We call this revised version DF As , because it can improve the performance when network is static, and call the original algorithm DF Am , because it applies when mobility is an issue. With this revised version, the total packet transmission time will be Ts + r × Tr + (R − r) × Tr0 X TDAT A + Toh ) = (R+1)(DIFS-SIFS) + BWi + (n + R)( n 1+r where Tr0 = DIF S − SIF S +

TDAT A n

+ Toh (11)

The simulation environment is set to consist of one WLAN network and one WPAN network in proximity. A Wi-Fi device and a BT device is assumed to be separated in less than 3 meters because such distance is considered the most severe interference scenario [3] [4]. This setup is quite common in office, home, airports, sports events, etc. For example, a PDA is connected to a laptop via Wi-Fi at rate 11Mbps, and concurrently a nearby master/slave pair, e.g. cellphone/headset, is communicating over Bluetooth link. The devices separation distance is intentionally assigned to reflect the specific interference scenario, which is whenever a BT device operates at Wi-Fi frequency band, if some Wi-Fi station is also active, then both transmissions would fail. However, if BT transmits outside the Wi-Fi frequency band, then both networks could transmit concurrently. In other words, the Pf is exactly 22/79 in our simulation. Also, each plotted value shown in each figure is the average results of at least 20 runs. Parameter Slot time DIFS SIFS PHY header MAC header Payload ACK CWmin CWmax

Assigned Value 20 µs 50 µs 10 µs 192 bits 224 bits 12000 bits 112 bits 32 1024

TABLE I 802.11 WLAN PARAMETERS Parameter TB TBA

Assigned Value 625 µs 366 µs

TABLE II B LUE T OOTH PARAMETERS

Without loss of the generality and in order to reduce the complexity of the simulation, the following assumptions have been made: (i) the packet payload length is fixed3 , along with other Wi-Fi and BT parameters shown in Tables I and II; (ii) like [8], RTS/CTS mechanism is not used, however, study shows the improvement remain with RTS/CTS; (iii) propagation delay is neglected; (iv) although there may be relative movement among devices, however, we assume the 3 However, if payload is variable, the algorithm could max F ragment length to achieve desire fragment length.

control

Fig. 5.

Throughput Improvement of the Wi-Fi with DFA

movement would have little affection on collision/interference pattern. The arrival rate of Wi-Fi packet is exponentially distributed. From Table I, we could calculate η, and for these Wi-Fi parameters the best η is two. Higher η would not further bring down κ of Eq. 9, but only increases overhead. For BT, both SCO and ACL traffic scenarios are simulated. In SCO link, the most popular HV3-type link is used, and a packet is generated every six time slots. A SCO packet is never retransmitted. If a BT slot is not reserved by SCO, the master could establish ACL link on per-slot basis. In ACL simulation, the DH1-type link is used such that one data packet occupies one BT time slot. The packet arrival rate of ACL is also exponentially distributed. Under different BT traffics, three cases are simulated: no fragmentation (legacy 802.11/state 1), fragmented Wi-Fi packets with mobile stations (DF Am /state 2m), and fragmented Wi-Fi packets with static stations (DF As /state 2s). Fig. 5 presents the improvement on the throughput of WiFi networks with DFA. When the PER of Wi-Fi reaches 0.5 or 0.6, DF Am will improve throughput by 15% and 30% respectively, and 28% and 56% respectively for DF As . One noticeable trend is that the throughput improvement grows exponentially with PER. The figure shows negative improvement for low PER. This is because for low PER, the gain from reducing retransmission caused by interference is lower than the increased overhead. However, no negative improvement would actually take place since our algorithm is dynamically adjusted based on the PER. When PER is below the threshold, there would be no state transition, thus nor deterioration would happen to Wi-Fi. In our results, throughput is defined as the ratio of time dedicated to payload to the total transmission time. Fig. 6 presents the Wi-Fi/BT throughput versus different WiFi/BT traffic load, where BT traffic is ACL data link. Fig. 6(a) indicates significant improvement on the throughput and it grows up with BT traffic load. The crossing point is controlled by background traffic. Intuitively, higher background traffic will trigger state transition earlier. Fig. 6(b) shows that slight

Fig. 8.

Average Wi-Fi Delay v.s. BT Traffic Load at τW iF i =0.5

improvement on BT throughput only happens when it is in high traffic load, but there is no negative effect whatsoever. Fig. 7 shows the Wi-Fi and BT Throughput versus different Wi-Fi traffic load when BT has two SCO links in presence. Since our mechanism is neither collaborative nor scheduling, we could foresee that our mechanism wouldn’t benefit from the recursive nature of SCO traffic. Fig. 7(a) indicates that even DFA provides improvement in low Wi-Fi traffic load, the improvement is only ordinary. Fig. 7(b) shows that the improvement on BT throughput is minor, but, however, still have some positive effect. Fig. 8 compares the average packet delay of Wi-Fi versus different BT traffic load. Average packet delay is simulated by transmitting a fixed length file, eg. 500 packets, and is calculated by dividing the total transmission time of the file by the number of packets. The improvement on delay comes from eliminating expensive retransmission caused by interference. Simulation results show that DFA could improve transmission delay significantly. V. C ONCLUSIONS We presented a non-collaborative mechanism, called Dynamic Fragmentation Algorithm (DFA), to improve the coexistence ability of Wi-Fi in the presence of BT inference. We begin with developing an analytical model which could faithfully reflect the interference phenomena between Wi-Fi and Bluetooth. According to the observation from our model, we proposed DFA to reinforce the coexistence ability of WiFi networks. With DFA, a Wi-Fi station could make state transition on a necessary basis to reduce interference and increase throughput. We also investigated the scenario of static network and came up with a revised model that can help DFA further improve the performance of Wi-Fi. In addition to the improvement on throughput, DFA also help Wi-Fi distinguish a collision from an interference. Thus, if the high PER isn’t caused by interference, DFA would recognize the difference and swiftly switch back to previous state to avoid the unnecessary overhead. The function of fragmentation has

(a) Wi-Fi Throughput v.s. BT Traffic Load at τW iF i =0.7 Fig. 6.

Throughput of the Wi-Fi and BT networks in the presence of ACL link.

(a) Wi-Fi Throughput v.s. Wi-Fi Traffic Load Fig. 7.

(b) BT Throughput v.s. Wi-Fi Traffic Load at τBT =0.7

(b) BT Throughput v.s. Wi-Fi Traffic Load

Throughput of the Wi-Fi and BT networks when there are two SCO links between BT master/slave.

already been built in IEEE 802.11, and thus our proposed solution can be easily implemented in a legacy 802.11. Simulation results show that our analytical model is very close to reality. With PER level higher than 0.6, we could get 56% improvement on throughput for static network, and 30% throughput improvement for mobile network. The improvements grow exponentially with PER. In the extensive simulation scenarios, Wi-Fi networks with DFA exhibit significant improvement on both throughput and delay. Since it’s a noncollaborative mechanism, the improvement on BT networks is not as good as that on Wi-Fi, but still has slight improvement. Although the current version of DFA has only two states and divides the packet into fragments with equal length, it is interesting to develop an extended model using more complicated fragmentation scheme, e.g. more states or advance partition methods. Our analytical model could provide clues for designing such advance partition mechanism and is also capable of evaluating attainable performance.

R EFERENCES [1] Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specification, ANSI/IEEE Std. 802.11, 1999. [2] Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specifications for Wireless Personal Area Networks (WPANs), IEEE Std. 802.15.1, 2002. [3] J. Lansford, A. Stephens, and R. Nevo, “Wi-fi (802.11b) and bluetooth: enabling coexistence,” IEEE Network, vol. 15, no. 5, pp. 20–27, Sep 2001. [4] N. Golmie, R. E. V. Dyck, A. Soltanian, A. Tonnerre, and O. Rebala, “Interference evaluation of bluetooth and ieee 802.11b systems,” ACM Wireless Networks, vol. 9, pp. 202–211, 2003. [5] IEEE 802.15 WPAN task group 2 (TG2). [Online]. Available: http://www.ieee802.org/15/pub/TG2-Coexistence-Mechanisms.html [6] N. Golmie, “Bluetooth dynamic scheduling and interference mitigation,” ACM Mobile Networks, vol. 9, no. 1, Feb 2004. [7] N. Golmie, N. Chevrollier, and O. Rebala, “Bluetooth and wlan coexistence: Challenges and solutions,” IEEE Wireless Communications Magazine, Dec 2003. [8] C. Chiasserini and R. Rao, “Coexistence Mechanisms for interference mitigation between IEEE 802.11 WLANs and bluetooth,” in Proc. of INFOCOM’02, vol. 2, Jun 2002, pp. 590–598. [9] I. Howitt and F. Awad, “Optimizing ieee 802.11b packet fragmentation in collocated bluetooth interference,” in IEEETransaction on Communications, vol. 53, no. 6, Jun 2005, pp. 936–938.

Coexistence Mechanism Using Dynamic Fragmentation ...

what proposed in [9] requires too much complexity to obtain the optimal solution ..... PDA is connected to a laptop via Wi-Fi at rate 11Mbps, and concurrently a ...

802KB Sizes 0 Downloads 196 Views

Recommend Documents

Dynamic Mechanism Design:
May 8, 2009 - Incentive Compatibility, Profit Maximization and Information Disclosure". Alessandro .... a solution to the profit%maximizing Relaxed Program.

Dynamic mechanism design: dynamic arrivals and ...
Jul 20, 2016 - both is a step towards realism and allows us to uncover new tradeoffs. The key properties of an ... When the horizon is infinite, and when the buyer arrives at each date with positive probability, the ... tax payers may face new suppli

Prepare Amplicon Libraries without Fragmentation Using the Ion ...
Mar 3, 2014 - This user bulletin describes how to prepare Ion libraries from ... depends on the specific Ion template preparation and sequencing kits used ...... For support visit lifetechnologies.com/support or email [email protected]

Coexistence in FemaleBonded Primate Groups
1994), interest in the social dynamics of femalebonded (FB) primates. 43 ... tive of most primate societies, FB groups are certainly much more common within the OldWorld .... females supported each other on only 10% of occasions (dropping to ...... r

Characterizing fragmentation in temperate South ...
processing we used the software ERDAS Imagine, Version. 8.2 (Leica .... compare landscapes of identical size, but it has also the disadvantage of ...... Monitoring environmental quality at the landscape scale. Bioscience 47 .... habitat networks.

Sheet reversing mechanism
Nov 25, 1974 - With the de?ector 32 in the intercept position, the sheets are fed into the nip of a ?rst roll pair formed by a drive roll 34 and an idler roll 36.

Localization Of License Plate Number Using Dynamic Image.pdf ...
There was a problem previewing this document. Retrying... Download. Connect more apps... Localization ... mic Image.pdf. Localization O ... amic Image.pdf.