Optics Communications 387 (2017) 290–295
Contents lists available at ScienceDirect
Optics Communications journal homepage: www.elsevier.com/locate/optcom
Capacity allocation mechanism based on diﬀerentiated QoS in 60 GHz radio-over-ﬁber local access network Yanbin Kou, Siming Liu, Weiheng Zhang, Guansheng Shen, Huiping Tian
State Key Laboratory of Information Photonics and Optical Communications, School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China
A R T I C L E I N F O
A BS T RAC T
Keywords: Medium access control (MAC) protocol Medium-transparent MAC (MT-MAC) Radio over ﬁber (RoF) Quality of service (QoS) Dynamic capacity allocation 60 GHz Local access network(LAN)
We present a dynamic capacity allocation mechanism based on the Quality of Service (QoS) for diﬀerent mobile users (MU) in 60 GHz radio-over-ﬁber (RoF) local access networks. The proposed mechanism is capable for collecting the request information of MUs to build a full list of MU capacity demands and service types at the Central Oﬃce (CO). A hybrid algorithm is introduced to implement the capacity allocation which can satisfy the requirements of diﬀerent MUs at diﬀerent network traﬃc loads. Compared with the weight dynamic frames assignment (WDFA) scheme, the Hybrid scheme can keep high priority MUs in low delay and maintain the packet loss rate less than 1% simultaneously. At the same time, low priority MUs have a relatively better performance.
1. Introduction With the unprecedented growth of mobile applications and the demands of higher data rate for mobile users (MUs), high capacity wireless communication networks are playing more and more important roles in today's society. By integrating the high capacity of optical access networks and the ﬂexibility of the wireless access networks, radio-over-ﬁber (RoF) technology can provide a higher data rate and can be more convenient in accessing to the MUs. It can be a promising technology for next generation broadband access networks [1,2]. RoF technology implementing the signal processing function at the Central Oﬃce (CO) and remote antenna units (RAUs), are only responsible for optical and radio frequency bidirectional conversions, resulting in low resource consumption and ﬂexible cell architectures. So the services of RAUs compacting and dynamically network capacity allocation are oﬀered at CO [3,4] which means the medium access control (MAC) protocols must be implemented in CO. However, RoF system introduces additional ﬁber propagation delay and impacts the performance of MAC protocols. For example, in IEEE802.11 and IEEE802.16. In order to solve this problem, the related researches have focused on changing existing protocol's MAC parameters to reduce negative eﬀects [5–9]. The medium-transparent MAC (MT-MAC) protocols are also proposed to allocate resources in both optical and wireless domains via a two-stage contention scheme for good communication performance [10–14]. However, with the rapid traﬃc growth of multimedia-based services, such as video, voice
and data service, the characteristics of the traﬃc are complex and various in access networks. So the diﬀerentiated QoS is a key issue to satisfy QoS of MUs in access network . In this paper, we propose a new scheme to satisfy the QoS requirements of diﬀerent MUs which based on the MT-MAC dynamic resource allocation mechanism. To meet the QoS requirements for diﬀerent MUs, a MU request table is used to save request information of MUs at the CO. Simulation results show that the mechanism can satisfy the performance demanding of high priority MUs while promising a better performance for low priority MUs compared with WDFA scheme at the same time. 2. The MT-MAC mechanism 2.1. Architecture of the RoF network In RoF system, the communication signal between CO and MUs is performed over a ﬁber based network and the 60 GHz wireless network, as is shown in Fig. 1. The signal processing and capacity allocation functions are located at the CO. And RAU modules are only responsible for the optical to RF and RF to optical conversion, without participate in actual signal processing. All uplink and downlink wavelengths are generated at CO and broadcast into the optical network. The downlink wavelength carries single-side-band (SSB) downlink data at a 60 GHz subcarrier from the CO to the RAU, while uplink wavelength carries uplink data back to the CO. Kalfas et al.
Corresponding author. E-mail address: [email protected]
http://dx.doi.org/10.1016/j.optcom.2016.11.043 Received 19 July 2016; Received in revised form 17 November 2016; Accepted 18 November 2016 Available online 01 December 2016 0030-4018/ © 2016 Elsevier B.V. All rights reserved.
Optics Communications 387 (2017) 290–295
Y. Kou et al.
Fig. 1. The structure of 60 GHz RoF network.
POLL, ID, and ACK packets. Every active MU chooses a random integer value y in the interval [0, m) corresponding the certain slot of RRF in which MU will try to register in the CO. The register procedure consists of require and response process. Firstly, the CO transmits a series of RRFs to the RAUs to identify the active MUs in the RAU cells. When receive the POLL packet from current slots, the active MUs will respond an ID packet to RAU. After the CO receives the ID packet of MU, it responds with an ACK packets, informing the MUs that it has been correctly identiﬁed. After the reception of the ACK packet, the MU will not participate in residual RRFs of the current SF. If two MUs choose the same y value to transmit their ID packets, it will be unavoidable for traﬃc collision. Therefore the CO will not respond with an ACK packet, forcing the two MUs to participate in the next RRF and regenerate a y value to send its ID packet. The CO continues to send RRFs until no collision occurs and all MUs are successfully registered. After those processes, the second contention period is completed, and the CO has fully acquired of the MUs that are active. If k RRFs are used to identify active MUs during the second contention period, (i-k) DFs will be broadcasted until the end of the SF. The CO initiates the DATA_TX period where the main exchange of data occurs in the form of a series of DFs transmissions until the end of the SF is reached.
provide more detailed description for the 60 GHz RoF physical layer architecture .
2.2. The MT-MAC protocol This section describes the main characteristics of the MT-MAC protocol's rule which are proposed to realize the medium arbitration process. A more detailed description of the protocol is provided in the Ref. . According to the protocol speciﬁcations, all traﬃc exchange is regulated by the CO. And the CO is attached to a number of RAUs through the ﬁber-based network which realizes optical to RF signal conversion and the communication with MUs. The capacity allocation is only negotiated by MUs and CO, without direct participating of RAUs. The data exchange between CO and MUs takes place in the optical and wireless domains, therefore we consider using two contention periods to allocate the capacity of network. The ﬁrst contention period is used to allocate optical wavelength pairs to RAU as MUs requested, and the second contention period allocates wireless resources to MUs. During the ﬁrst contention period, the control wavelength λ c carries the beacon pulse emitting from CO to RAUs then propagates to target MUs in their respective ranges. When target MU detects the pulse, the MU will respond immediately by emitting a short pulse in the same duration to inform the CO its presence and its data transmission requirement, otherwise the MU will keep silent. After receiving the response pulse, the CO acquires the communication acquirements of MUs in the RAUs and assigns data wavelength pairs to the RAUs. Then the ﬁrst contention period is completed. When the data wavelength pairs are less than active RAUs quantity, the CO assigns the data wavelength pairs in a round-robin fashion for all RAUs. The second contention period is entirely applied to the wireless part of RoF network, where traﬃc exchange is divided into SuperFrames (SFs), as shown in Fig. 2. Each SF contains a number of frames including Resource Requesting Frames (RRF) and Data Frames (DF). Each RRF has a constant number of slots, m. And each slot comprises
3. The proposed approach based on QoS In this section, in order to satisfy the QoS requirements of MUs in diﬀerent priorities, we classify the MUs into multiple classes of services (CoS) – CoS0 (high priority), CoS1 (middle priority), CoS2 (low priority). At CO side, there are plenty of modules such as MU request table, multi-queues and scheduler, as shown in Fig. 3. The MU request table is introduced to save uplink request information of MUs. Multiqueues buﬀer the downlink data packets which are designated by the service type of packet. In order to acquire the request information of MUs, the RRFs and DFs in MT-MAC protocol are modiﬁed to satisfy the requirement of QoS.
Fig. 2. The structure of SuperFrames.
Optics Communications 387 (2017) 290–295
Y. Kou et al.
Fig. 3. The framework of CO supporting QoS.
3.1. The modiﬁed RRFs
During the second contention, the ID packets of RRFs are modiﬁed containing identify number of MUs, CoS and request number of bytes (NoB), which are used to represent the request information of MUs. When the CO receives the ID packets, the CO adds request information of MUs to the MU request table.
∑j Bik, j−1 ∑i, j Bik, j−1
(L− ∑ Lik, RRFs ) (1)
is the number where L is the total number of DFs and RRFs, and of RRFs for the RAU i in the cycle k. The DFA allocates the numbers of frames to each RAU in order to be shared by all MUs. When the traﬃc load is high, the allocated frame sizes of CoS0 MUs could not satisfy the QoS requirement. For this reason, the WDFA is introduced. A weight wj is assigned to each CoS and then Eq. (1) is employed. The number of DFs assigning to RAU i in cycle k is shown in Eq. (2).
3.2. The modiﬁed DFs During the DATA_TX period, The DFs consist of POLL, DATA, ACK and REPORT packets. The DATA packet is modiﬁed with an additional 2-bit ﬁeld to deﬁne CoS, which is used to represent priorities for traﬃc. The REPORT packet modiﬁcations must be endowed with the CoS and NoB of MUs, which are used to represent the MU request information in the next poll cycle. The POLL packets of DFs contain the identify numbers and assign transmission opportunity windows (TX_OP) of MUs in current poll cycle. It will inform the relative MU transmits data in the certain slot of current poll cycle. After matching the ID, the MU sends DATA packets to the CO through RAU. The RAU only realize RFoptical conversion, without employing any advance signal process function. And before the end of the MU sending frame sizes, the MU will send a REPORT packet to inform CO about its traﬃc request in the next poll cycle if it needs to transmit data, and then the CO updates the MU request table. On the contrary, if the MU does not need information transmission, it will be removed and do not participate in capacity allocation in next poll cycle (SuperFrame). The CO will not respond an ACK packet immediately after the reception of a data packet. The corresponding ACK is piggybacked in a subsequent POLL packet .
∑j wj × Bik, j−1 ∑i, j wj × Bik, j−1
(L− ∑ Lik, RRFs ) i
The WDFA mechanism can satisfy the QoS requirements of diﬀerent MUs by distributing rational weights for diﬀerent CoSs. RAUs with higher priority MUs get a relatively greater number of DFs even if the actual requests traﬃc of MU is low. At the same time, it's diﬃcult to select a single optimal set of weights when the actual traﬃc patterns for diﬀerent CoSs are expected to vary over time. Unfairness may appear if there is a signiﬁcant discrepancy in MUs with diﬀerent priorities mix among diﬀerent RAUs. For example, a RAU only contains MUs in high priority, which may getting more DFs causing waste, however, other RAUs can’t acquire enough DFs. In order to improve fairness between the high and middle/low priority-MUs, we propose a hybrid method of operation which allows automatically switching between an adaptively weight DFA (AWDFA) and a DFA mechanism. The latter provides the same weights for all MUs and the middle and low priority ones will have better performance than in AWDFA scheme. In the hybrid scheme, the number of DFs in the network is divided into two dynamically re-organized groups. These two groups are controlled by AWDFA and DFA algorithms, respectively. We introduce a new parameter called frames grouping ratio k to classify the frames groups. The frames grouping ratio deﬁnes the ratio of frames being used by the DFA algorithm over the total number of available frames. It means the value of frames grouping ratio is 0 while all frames are allocated using AWDFA. Instead, a value of 1 denotes that frames allocation is totally based on DFA. Dynamic capacity allocation is implemented in this scheme by automatically adjusting the frames grouping ratio value according to the actual traﬃc condition. The cycle k of frames grouping ratio is calculated by exploiting the diﬀerence in required bandwidth between consecutive cycles, for each priority. If the required bandwidth in cycle k for the high priority is greater than in polling cycle k−1, the frames grouping ratio will be reduced and the number of DFs for AWDFA will increases. The scheme realizes similarly for middle and low priorities, which are shown in the
3.3. Analysis of the capacity allocation mechanism After second contention period, the CO acquires the uplink request information of MUs by the MU request table. At the same time, in the CO buﬀer queue, downlink packets are also diﬀerentiated based on the CoS. It must be ensured that RoF capacity allocation mechanism takes the QoS requirements of the MUs. According to the MU request information, the dynamic capacity allocation process assigns DFs for the MUs. In the ﬁx poll cycle T, the sum of RRFs and DFs is always constant and equal to L. If M RRFs are used for the 2nd contention period, (L-M) DFs will be broadcasted. The dynamic frames assignment (DFA) algorithm process is shown as follows. Bik, j−1 denotes the number of bytes reported by RAU i for MU CoS j in cycle k-1. The number of DFs assigned to RAU i in cycle k will be 292
Optics Communications 387 (2017) 290–295
Y. Kou et al.
Table 1 Simulation Parameters.
k k −1 ⎧ k −1 ( ∑ B k > ∑i Bik, j−1 ⎪ (1 − α )∙ =0 or ∑i Bi, j =1,2 < ∑i Bi, j =1,2 ) i i, j =0 k=⎨ k k −1 ⎪ k −1 ( ∑ B k < ∑i Bik, j−1 =0 or ∑i Bi, j =1,2 > ∑i Bi, j =1,2 ) ⎩ (1 + α )∙ i i, j =0
(3) where α is the quantized adaptation step of the frames grouping ratio parameter with the changing traﬃc. As a result, the number of assigned DFs for using AWDFA and DFA algorithms can be expressed.
⎤ ⎡⎛ ⎞ k LDFA =⎢ ⎜⎜L− ∑ Lik, RRFs⎟⎟ × k⎥ ⎥⎦ ⎢⎣ ⎝ ⎠ i
⎤ ⎡⎛ ⎞ k LAWDFA =⎢ ⎜⎜L− ∑ Lik, RRFs⎟⎟ ×(1 − k) ⎥ ⎢⎣ ⎝ ⎠ ⎦⎥ i
where β is the quantized weight adaptation step for each RAU with the changing traﬃc. Therefore, the number of DFs assigned to RAU i in cycle k is expressed:
∑j wik, j × B
k −1 i, j + Lk k −1 AWDFA
∑i, j wik, j × B
∑j Lik, j−1 ∑i, j Lik, j−1
Fiber propagation delay Air propagation delay Slots in RRF DATA size Data bitrate POLL, ID, REPORT size ACK size Buﬀer size Poll cycle The proportion of traﬃc for CoS0, CoS1, CoS2
1 μs/200 m 0.16 μs 10 64–1500bytes 1 Gbps 64bytes 8bytes 80 pack 2 ms 2:4:4
Fig. 4 displays the end to end packet delay for all CoS versus with load changing, ranging from 20% up to 100%. For the DFA scheme, the packet delay of each CoS is the same since the capacity allocation only depends on their request bandwidth. In Fig. 4(a), when the network load reaches 1.0, the delay of CoS0 is less than 4 ms and 3 ms in Hybrid scheme and WDFA scheme, respectively. At the same time, the delay of CoS2 in the two schemes is 11 ms and 8.6 ms, as shown in Fig. 4(c). The reason is that the high priority traﬃc, CoS0, has greater weight compared to other priorities in use of the WDFA scheme. And the Hybrid scheme combines DFA and AWDFA algorithms which can obtain more reasonable distribution of DFs among priority MUs. Fig. 5 shows the packet loss rate performance for three priorities. The packet loss rate is deﬁned as the ratio of the number of packets discarded over the total packets generated for each CoS. Fig. 5(a) displays the packet loss rate of CoS0 by using diﬀerent schemes. When the traﬃc load is less than 0.7, all CoS0 packets loss rates are zero. The packet loss rate of the DFA scheme rises quickly when the load is varying from 0.8 to 1.0, and the packet loss rate reaches 17% when the network load reach 1.0. Meantime the packet loss rate is less than 1% in Hybrid scheme and in WDFA scheme. The packet loss rate of CoS1 is shown in Fig. 5(b). The packet loss rate of Hybrid scheme and WDFA scheme decreases signiﬁcantly compared with DFA scheme. Fig. 5(c) shows the packet loss rate of CoS2 using in the diﬀerent schemes, when the network load is 1.0, the Hybrid scheme packet loss rate is declined by 11% compared with WDFA scheme. At the same time, the packet loss rate of CoS0 is less than 1%, as is shown in Fig. 5(a). This is due to the fact that the Hybrid scheme can adjust the frames grouping ratio and the weights of RAU with traﬃc changing. It can guarantee the QoS requirements of MUs in high priority and assign more DFs for MUs in low priority. Fig. 6(a), (b) and (c) display the throughput performance of three priorities MUs (CoS0, CoS1 and CoS2) respectively, for diﬀerent schemes. For the DFA scheme, when the traﬃc load is less than 0.8, the throughput of each CoS is increasing gradually with the load increasing. The trend of the throughput increase is almost linear. This is due to the packet loss rate almost zero, as shown in the Fig. 5. When the load continues increasing, the packet loss rate of all CoS rises quickly, so the throughput for each CoS will no longer change. For the WDFA scheme, the throughput of CoS0 will keep increasing linearly, even the traﬃc load up to 1.0. However, the throughput of CoS2 is quickly decreased when the traﬃc load is more than 0.8. Compared with the WDFA scheme, the Hybrid scheme almost has the same throughput performances for CoS0 and CoS1, and the CoS2 performance has been signiﬁcantly improved. It is because WDFA algorithm allocates more DFs for CoS0. So the packet loss rate of COS0 still keep zero with traﬃc load increasing, as shown in Fig. 5(a). When the traﬃc load is high, it will sacriﬁce the low priority MUs to ensure the high priority UEs having enough DFs, even if the actual request DFs of CoS0 is less than assign DFs. The Hybrid scheme can decrease DFs waste by adjusting the parameter of frames grouping ratio and weights of RAU with traﬃc changing. The packet loss rate of CoS2 decreases signiﬁ-
For each RAU, frames allocation is further enhanced in the proposed algorithm by means of adaptable weights. The MUs in diﬀerent priorities will get optimize frames size within the respective RAU. The weight distribution for the cycle k is performed based on the priority and weight adjustment process for each RAU, and the priority is shown in the following equation: k −1 ⎧ (Bik, j < Bik, j−1 ) ⎪ (1 − β )∙wi, j wik, j =⎨ k −1 ⎪ (Bik, j > Bik, j−1 ) ⎩ (1 + β )∙wi, j
4. Performance evaluation To verify the validity of the proposed MT-MAC dynamic capacity allocation mechanism based on QoS in 60 GHz RoF over bus network, we adopt the OPNET model to produce and value our proposed mechanism. In the simulations, we just consider a ﬁxed number of 3 RAUs while each RAU serves 10 MUs. The distance between the ﬁrst RAU and CO equals to 2.5 km and the ﬁber intervals amongst the RAU modules are 200 m. In addition, the channel is considered to operate under ideal conditions and all users are equipped with buﬀers to accommodate generated traﬃc. A typical ﬁxed, 2ms polling cycle is used to periodically allocate resources to the MUs and the MUs are used for transmitting data. It means that the number of DFs assigned to MU's in current poll cycle is on the basis of the request information of MUs. In order to satisfy the QoS requirements of diﬀerent MUs for diﬀerent services, we divided generate packets of the MUs to high priority, middle priority and low priority. These priorities occupied 20%, 40%, 40% of the total generated packets. We use the ON/OFF source model to generate data packets. The packet size was uniformly generated between 64 and 1500 Bytes . The ON/OFF time exhibits an exponential distribution and the packet inter-arrival time within ON periods is implemented by a Pareto distribution to generate burst traﬃc. Table 1 shows the relevant simulation parameters. Three simulation scenarios are evaluated to compare between DFA scheme, WDFA scheme and hybrid scheme. With respect to WDFA, the weight wj of high, middle and low priority is 10, 5 and 2 respectively. For hybrid algorithm at the beginning of the simulation, the initial frames grouping ratio k and weight wik, j can be set to 0.5 and 10:5:2, respectively. The quantized adaptation step of both frames grouping ratio α and weight of RAU β are set to 0.01. 293
Optics Communications 387 (2017) 290–295
Y. Kou et al.
Fig. 4. End-to-end packet delay of (a) CoS0 (b) CoS1 (c) CoS2.
Fig. 5. Packet loss rate of (a) CoS0 (b) CoS1 (c) CoS2.
Fig. 6. Throughput of (a) CoS0 (b) CoS1 (c) CoS2.
cantly compared with WDFA scheme, as shown in Fig. 5(c). Fig. 7 shows the delay jitter performance for three priorities. The N jitter is calculated as δ 2= ∑1 (di − D )2 / N , where di is the delay time of packet i, N is the total number of received packets, and D is the mean packet delay . When the traﬃc load exceeds 0.8, the jitter of DFA scheme will rise quickly, as shown in Fig. 7(a). The use of WDFA scheme is shown in Fig. 7(b). When the load exceeds 0.7, the CoS2 jitter is quickly increasing with traﬃc load increasing, and the jitter reaches 13.7 when the network load is 1.0, while the CoS0 jitter is still keeping low. Fig. 7(c) shows the mean jitter of each CoS used in the Hybrid scheme. When the network load is 1.0, the CoS2 mean jitter signiﬁcantly declined compared with WDFA scheme. At the same time, the mean jitter of CoS0 is keeping low.
In this paper, we propose a dynamic capacity allocation mechanism based on the QoS for 60-GHz RoF local access network. Based on the QoS of diﬀerent MUs, the MUs’ traﬃc requests are divided into multiple priorities. The CO acquires MUs’ traﬃc requests and traﬃc priorities. We further analysis and simulate the delay, packet loss rate and throughput of DFA scheme, WDFA scheme and Hybrid scheme, respectively. Compared with WDFA scheme, the Hybrid scheme can keep the CoS0 with low delay and a packet loss rate less than 1%, while the CoS2 having a relative better performance at the same time.
Optics Communications 387 (2017) 290–295
Y. Kou et al.
Fig. 7. Jitter with diﬀerent traﬃc loads (a) DFA (b) WDFA (c) Hybrid scheme. Am. Trans. 8 (4) (2010) 377–384.  S.Deronne, V.Moeyaert, S.Bette,Analysis of the MAC Performances in 802.11 g Radio-Over-ﬁber systems, in Communications and Vehicular Technology in the Benelux (SCVT), 2011  G. Kalfas, N. Pleros, An agile and medium-transparent MAC protocol for 60 GHz radio-over-ﬁber local access networks, J. Light. Technol. 28 (16) (2010) 2315–2326.  G. Kalfas, N. Pleros, K. Tsagkaris, L. Alonso, C. Verikoukis, Saturation throughput performance analysis of a medium transparent MAC protocol for 60 GHz radioover-ﬁber networks, J. Light. Technol. 29 (24) (2011) 3777–3785.  P.Maniotis, G.Kalfas, L.Alonso, C.Verikoukis, N.Pleros Throughput and delay fairness through an agile medium transparent MAC protocol for 60 GHz ﬁberwireless LAN networks, in: Proceedings of the Hybrid Optical-Wireless Access Networks Workshop (WCNC), Paris, France, 2012.  Z. Xu, H. Wang, Y. Ji, Multichannel resource allocation mechanism for 60 GHz radio-over-ﬁber local access networks, J. Opt. Commun. Netw. 5 (3) (2013) 254–260.  G. Kalfas, P. Maniotis, S. Markou, D. Tsiokos, N. Pleros, L. Alonso, C. Verikoukis, Client-weighted medium-transparent mac protocol for user-centric fairness in 60 ghz radio-over-ﬁber WLANs, J. Light. Technol. 6 (1) (2014) 33–44.  Jijun Zhao, Nam Feng, Danping Ren, Research on energy eﬃciency based on diﬀerentiated QoS in ﬁber-wireless broadband access network, Optik 126 (2015) 350–355.  Wansu Lim, Pandelis Kourtessis, Konstantinos Kanonakis, Milos Milosavljevic, Ioannis Tomkos, John M. Senior, Dynamic bandwidth allocation in heterogeneous ofdma-pons featuring intelligent LTE-A traﬃc queuing, J. Light. Technol. 32 (10) (2014) 1877–1885.  I.-Shyan Hwang, AliAkbar Nikoukar, Ku-Chieh Chen, Andrew Tanny Liem, ChingHu Lu, QoS Enhancement of live IPTV using an extended real-time streaming protocol in ethernet passive optical Networks, J. Opt. Commun. Netw. 6 (8) (2014) 695–704.
Acknowledgements This work was supported in part by the NSFC under Grant No. 61431003, and 61372038 and the Fund of the State Key Laboratory of Information Photonics and Optical Communications (Beijing University of Post and Telecommunications), China. References  N. Ghazisaidi, M. Maier, Fiber-wireless (FiWi) access networks: challenges and opportunities, IEEE Netw. 25 (2011) 36–42.  H.T. Win, A.S. Khan, Pathan, on the issues and challenges of ﬁber-wireless (FiWi), Netw. J. Eng. (2013) 1–11.  B. Masella, X. Zhang, A novel single wavelength balanced system for radio over ﬁber links, IEEE Photonics Technol. Lett. 18 (1) (2006) 301–303.  T. Kuri, T. Nakasyotani, H. Toda, K. Kitayama, Characterizations of supercontinuum light source for WDM millimeter-wave-band radio-on-ﬁber systems, IEEE Photonics Technol. Lett. 17 (6) (2005) 1274–1276.  A. Das, M. Mjeku, A. Nkansah, N.J. Gomes, Eﬀects on IEEE 802.11 MAC throughput in wireless LAN over ﬁber systems, J. Light. Technol. 25 (11) (2007) 3321–3328.  M. Mjeku, N.J. Gomes, Analysis of the request to send/clear to send exchange in WLAN over ﬁber networks, J. Light. Technol. 26 (15) (2008) 2531–2539.  B. Kalantari-Sabet, M. Mjeku, N.J. Gomes, J.E. Mitchell, Performance impairments in single-mode radio-over-ﬁber systems due to MAC constraints, J. Light. Technol. 26 (15) (2008) 2540–2548.  P.H. Gomes, N.L.S. da Fonseca, O.C. Branquinho, Analysis of performance degradation in radio-over-ﬁber systems based on IEEE 802.16 protocol, IEEE Lat.