Model validation

We validate the derivation of successful transmission probability with simulation results. we assume CR nodes are always backlogged. In the simulation experiment, the mean and standard deviation of PUs' traffic load are 0.5 and 0.1, respectively. The comparison is summarized in Table I. There exists discrepancy between the derivation and simulation results, which is due to the setting of standard deviation. In our derivation, a CR node uses the mean traffic load value of PUs to estimate the corresponding successful transmission probability. However, in simulation experiment, channel idle time maybe cannot accommodate q_i frames, i=1, 2. In such a situation, PUs should wait for transmission completion. Those events are not counted in the calculation of the successful transmission probability. Thus the successful probability of simulation result is smaller than that of derivation. One significant achievement of our mechanism is that CR nodes utilize at least 70% of the channel idle time.

IV. PERFORMANCE EVALUATION

In this section, we develop a simulation program to compare the performance of the designed CACS mechanism

with OSA-MAC [2], SSA-MAC [3], CH-MAC [4], and DRA- MAC [5].

In this experiment, there are one control channel, and five data channel. The PU traffic load on data channel

is poisson distribution with rate Moreover,

we set and CR

nodes are always backlogged. The bandwidth of a data channel is 2 Mbps. Frame size is 2048 bytes. The transmission ranges of PUs, CR nodes, and CR APs are 150 meters, 100 meters, and 100 meters, respectively. The duration of DIFS and SIFS is 0.05 and 0.01 ms, accordingly. For SSA-MAC, the settings of TXQ and RTV are 4 and 1, respectively. The simulation time is 100 seconds. The observed performance metrics include

"utilization of channel idle time", and "average tries of channel search".

We first investigate the utilization of channel idle time of various mechanisms, and the results are shown in Fig. 4. We found that CSA-MAC (with handshaking) performs better than other MAC protocols. The reasons have twofold: setting transmission quota according to PUs' traffic load; and adapting channel sensing time based on measured channel quality. As a result, CR nodes utilize channel idle time as much as possible.

The performance gap between CSA-MAC with handshaking and without handshaking is caused by different dwell time when sensing a busy channel. Indeed, the dwell time for CSA-

Figure 4. The utilization of channel idle time v.s. the number of CR pairs.

Figure 5. The average tries of channel search v.s. the number of CR

MAC with handshaking is , while it's for ^pairs

CSA-MAC without handshaking. The reason of low utilization for OSA-MAC is that a CR pair only exchanges one data frame when occupying a data channel. Moreover, the common drawback of DRA-MAC and CH-MAC is that if being aware of PU presence on the sensed data channel, CR nodes will stay at that channel for five slots, thus resulting in low utilization.

SSA-MAC has a mechanism for PUs to interrupt CR transmission. Thus, SSA-MAC performs worse than CSA- MAC (with handshaking).

Next, the performance of the average tries of channel search for various mechanisms is in Fig. 5. It is common for all mechanisms that, when the number of CR pairs increases, the average tries of channel search also increases. Besides, CSA- MAC (with handshaking) outperforms CH-MAC and DRA- MAC. The reason is, in CH-MAC and DRA-MAC, a CR sender does not select channels according to PUs' traffic loads, and thus may frequently sense busy channels. Besides, comparing with random hopping sequence performed in SSA- MAC, our estimation of successful transmission probability makes a great impact when there are more than five CR pairs.

In OSA-MAC, a CR sender only sense once during a fixed

transmission probability, sensing time, and transmission quota, of each data channel. The three parameters are derived through an analytical queueing model, and the support of powerful cloud servers. Two versions of CSA-MAC are presented and compared in this paper, with handshaking and without handshaking. The simulation results showed that CSA-MAC with handshaking performs better in the utilization of channel idle time, while CSA-MAC without handshaking diminishes the average tries of channel search. In the future, we will investigate the impact of different arrival rate of CR users and extend this work to multi-hop CR flows.

ACKNOWLEDGEMENT

This work was supported in part by National Science Council under grants NSC 99-3113-P-009-004 and NSC 100- 2219-E-009-005, and in part by the Information and Communications Research Laboratories (ICL), Industrial Technology Research Institute (ITRI), Taiwan, under grant A352BW2100.

REFERENCES

period. Thus, OSA-MAC has the least tries of channel search among all mechanisms, while its drawback is low utilization of channel idle time as previously discussed. Note that CSA (without handshaking) still performs better than most compared protocols. The reason is that a CR pair has to wait

[1]

[2]

FCC Spectrum Policy Task Force, "Report of the Spectrum Efficiency Working Group," FCC, Technical Report 02-155, 2002.

L. Long and E. Hossain, "OSA-MAC: A MAC Protocol for Opportunistic Spectrum Access in Cognitive Radio Networks," in Proceedings of IEEE WCNC, 2008.

for when sensing a busy channel, which implies that ^[3] T.C. Lou, H.L Chao, and T.J Tsai, "On Synchronized Channel Sensing and Accessing for Cognitive Radio Users in IEEE 802.11 Wireless

CSA-MAC (without handshaking) has relative long sensing time.

V. CONCLUSIONS

In this paper, we proposed a cloud server-assisted MAC protocol, named CSA-MAC, for infrastructure-based cognitive radio networks. In CSA-MAC, each CR nodes maintains a channel preference matrix, which records the successful

[4]

[5]

Networks," in Proc of IEEE PIMRC, 2009.

H. Su, and X. Zhang, "Channel-hopping Based Single Transceiver MAC for Cognitive Radio Networks," in Proc. of IEEE CISS, 2008.

L. Jiao, and F.-Y. Li, "A Single Radio Based Channel Datarate-aware Parallel Rendezvous MAC Protocol for Cognitive Radio Networks," in Proc.

of IEEE LCN, 2009.

[6]

[7]

[8]

[9]

IEEE 802 LAN/MAN Standards Committee 802.22 WG on WRANs.

http://www.ieee802.org/22.

D-H. Huang, S.-H. Wu, and P.-H. Wang, "Cooperative Spectrum

Sensing and Locationing: A Sparse Bayesian Learning Approach," in Proc. of IEEE GLOBECOM, 2010.

W-Y Lee, and I.F. Akyildiz, "Optimal Spectrum Sensing Framework for Cognitive Radio Network," IEEE Wireless Communications, vol. 7, pp. 3845-3857, Oct. 2008.

Kleinrock, L., "Queueing systems. Vol. II: Computer applications," John.

Wiley & Sons, New York, 1976.