Open Access
Add to BibSonomyAbstract
Elastic optical network (EON) has been proposed recently as a spectrum-efficient optical layer to adapt to rapidly-increasing traffic demands instead of current deployed wavelength-division-multiplexing (WDM) optical network. In contrast with conventional WDM optical cross-connect (OXCs) based on wavelength selective switches (WSSs), the EON OXCs are based on spectrum selective switches (SSSs) which are much more expensive than WSSs, especially for large-scale switching architectures. So the transition cost from WDM OXCs to EON OXCs is a major obstacle to realizing EON. In this paper, we propose and experimentally demonstrate a transition OXC (TOXC) structure based on 2-stage cascading switching architectures, which make full use of available WSSs in current deployed WDM OXCs to reduce number and port count of required SSSs. Moreover, we propose a contention-aware spectrum allocation (CASA) scheme for EON built with the proposed TOXCs. We show by simulation that the TOXCs reduce the network capital expenditure transiting from WDM optical network to EON about 50%, with a minor traffic blocking performance degradation and about 10% accommodated traffic number detriment compared with all-SSS EON OXC architectures.
© 2016 Optical Society of America
1. Introduction
With the explosive growth of Internet traffic of high burstiness and varying size, optical networks should support both huge network traffic accommodation and flexible network resource operation. Empowered by technologies like O-OFDM [1–3 ] and Nyquist WDM [4,5 ], elastic optical networks (EONs) [6,7 ] have been introduced recently to cope with the high-capacity traffic delivery and variable bandwidth requirements. More specifically, in EONs, continuous spectrum slots with finer granularities can be constructed as a superchannel, which can be assigned to variable-size traffic (i.e., from sub-wavelength to super-wavelength). Enabling technologies, including bandwidth variable transponder, spectrum selective switches (SSSs) and coherent detection, have been experimentally demonstrated in previous studies [8–10 ].
The optical cross-connect (OXC) is the key switching element in both wavelength-division-multiplexing (WDM) networks and EONs. In conventional WDM networks, the implementation of OXC is based on the wavelength selective switch (WSS), which has coarse grids with wavelength spacing of 50 GHz or 100 GHz. WSSs and optical couplers/splitters are cascaded in the conventional WDM OXC to support the switching function. Generally, following the same way, EON OXCs could be constructed based on SSSs with much finer spectrum grids (e.g. 12.5 GHz or less).
However, with the over 40% increase of traffic volume per year [11] and the extension of network connection dimensions [12,13 ], OXC connection degrees are expected to expand significantly which requires expensive larger port count WSSs/SSSs. As the capital expenditure (CAPEX) seriously constrains the future network infrastructure upgrades, service providers need to construct more cost-efficient and scalable networking infrastructures. Some studies have been carried out to reduce the CAPEX of large-scale OXCs in WDM networks. For example, the interconnected subsystem OXC [14] which consists of several inter-connected broadcast-and-select subsystems has been implemented to reduce the cost of inter-layer switching in multi-fiber networks. Hybrid OXCs based on WSSs, photonic switches (PXCs) and/or waveband selective switches (WBSSs) [15,16 ] have adopted similar design to reduce the inter-fiber-layer cost. In these designs, cost efficient PXCs and WBSSs have been used to construct inter-layer switching architectures. Besides, hierarchical OXCs [17–20 ] have been introduced by employing optical devices with different spectral granularities.
Furthermore, to evolve OXCs from WDM to EON, the cost gap between SSSs and WSSs should be considered as well. At present, 100 GHz-grid WSSs are based on PLC (planar-lightwave-circuit) or MEMS (micro-electro-mechanical system) technology while 12.5 GHz SSSs are mostly based on LCoS (Liquid Crystal on Silicon) technology which is more expensive. So, due to the high cost and fabrication difficulty of SSS [13], the port count of SSS is seriously restricted and has been the bottleneck to realize EONs.
In this paper, we focus on reducing the transition cost from WDM networks to EONs. Although the SSSs are irreplaceable in EON OXCs due to the fine-grid switching operation, we reduce the required port count of SSSs by using WSSs to construct the OXC with the same switching scale. We propose a novel transition OXC (TOXC) with a two-stage switching architecture: the former stage is based on port-count-limited SSSs while the latter stage reuses WSSs in current deployed WDM OXCs. Then, we experimentally demonstrate the two-stage switching based on polarization-division-multiplexed (PDM) OFDM superchannels. Also, we propose a contention-aware spectrum allocation (CASA) scheme for the TOXCs to reduce the conflicts caused by limited flex-grid switching capability. The simulation results show that, the proposed TOXCs significantly reduce the CAPEX with only a slight decrease of bandwidth blocking probability (BBP) performance and network accommodation (NA) compared with the high-cost all-SSS EON OXCs.
The rest of paper is organized as follows. The proposed transition strategy of OXC architecture from WDM to EON is introduced in section II. We introduce our experimental demonstration and results in section III. The proposed RSA for the TOXC architecture is illustrated in section IV. Simulation results of proposed TOXC and RSA scheme are presented in section V. Section VI concludes the paper.
2. Transition strategy of OXC architecture from WDM to EON
In this section, we introduce the transition OXC (TOXC) architecture and the strategies. Figure 1 (a) and 1(d) show the conventional WDM OXC and all-SSS EON OXC with port count of n respectively. As we can see, fully-connected broadcast-and-select switching structures is employed and n 1 × n SSSs are needed to construct an all-SSS EON OXC (Fig. 1(d)). For a conventional EON OXC, the SSSs provide both spectrum selective and switching functions. The former function relies on the spectrum granularity of SSSs while the later function depends on the switching scale/port count of SSSs. To reduce the required switching scale/port count of SSS but keep the fine granularity switching function, the proposed TOXCs isolate these two functions into two stages and use cost-effective WSSs instead of SSSs for the switching function. As shown in Fig. 1(b) and 1(c), the two stages can be described as follows:
Fig. 1 OXC Architectures: (a) WDM OXC; (b) proposed CTS-TOXC; (c) proposed MS-TOXC; (d) all-SSS EON OXC.
- 1) Spectrum selective stage: For each OXC input port, a 1 × c fine-granular SSS separates spectral adjacent channels and distributes them to different WSSs in the next stage (c = 2 in Fig. 1(b) and 1(c)).
- 2) Switching stage: The channels from the previous stage is switched by coarse WDM-grid WSSs and then combined by optical couplers.
The proposed TOXC can make full use of WSSs in deployed WDM OXC with SSS and WSS cascading structure. It may have two forms, as shown in Fig. 1(b) and 1(c). Figure 1(b) shows a cascading two-stage TOXC (CTS-TOXC), in which n 1 × 2 SSSs are used in the first stage and two n × n WDM-OXCs are used in the second stage. Figure 1(c) shows a mixed-stage TOXC (MS-TOXC), in which n/2 1 × n SSSs are used for half of the OXC ports, while n/2 1 × 2 SSSs and a WDM-OXC are used for the rest.
It should be noted that c is a constant for TOXCs. For each WDM-grid, the number of channels (NoC) can be larger than c but the number of their directions (NoD) has to be no more than c. When NoD ≤ c, the 1 × c SSS can handle the requests regardless of NoC. A switching example of TOXC is shown in Fig. 1(b). The bandwidth of each spectrum slot is 12.5 GHz, and each WDM-grid contains 8 spectrum slots (including a guardband). The guardband is set between adjacent WDM-grids due to the filtering effect of WSSs to ensure the transmission quality. This may lead to extra spectrum wastage. There are 5 switched channels coming from the OXC input-port #n/2 + 1, as shown in the example:
Incoming: channels with different directions are marked in different colors. Channel A (orange) occupies the spectrum slot 1; channel B (blue) occupies the spectrum slots 3 ~5; channel C (orange) occupies the spectrum slot 7 and has the same direction (outgoing port) as channel A; channel D (green) occupies the spectrum slot 9 ~12; channel E (red) occupies the spectrum slots 14 ~15. The spectrum slots 0, 2, 6, 8, 13, and 16 are guardbands.
Spectrum selective stage (SSS #n/2 + 1): The spectral adjacent channels in the same WDM grid are separated from each other and transported to different WSSs: channels A, C, and E to WSS #1; channels B and D to WSS #2.
Switching stage (WDM OXC #2): the channels are switched by WDM-grid switching architectures and transported to different OXC ports: output-port # 1, #n/2, #n/2 + 1, and #n.
It should be noted that, generally, the large-port-count SSS is implemented with cascaded small port count SSS, so the cost of a SSS can be evaluated by its port count. The proposed OXCs can effectively reduce the required port count of SSSs. For the MS-TOXC, the overall port count is n 2/2 + n (a 1 × n SSS is counted as n ports; there are n/2 1 × n SSSs and n/2 1 × 2 SSSs), while that is 2n for the CTS-TOXC (with an extra n × n WDM OXC). The port counts of both OXC architectures are much fewer than n 2 in the all-SSS EON OXC architecture [7].
3. Experimental demonstration of TOXC architectures
In this section, we present a proof-of-principle experimental demonstration of CTS-TOXC switching on optical superchannels. The experimental setup is depicted in Fig. 2 :
Fig. 2 Experimental setup of superchannel switching by TOXCs (insets: superchannels and filtering effects of SSSs and WSSs).
Transmitter (Tx): In the Tx, three lasers with the frequency spacing of 12.5 GHz are coupled by cascaded polarization maintaining couplers (PMCs). Arbitrary waveform generator (Tektronix AWG7122B) operating at 12 GS/s is used to generate the baseband OFDM signals. 256 subcarriers are adopted, among which there are 200 data subcarriers and 4 pilot subcarriers and the rest are filled with zero as guardband. The oversampling ratio is 256/204 = 1.2549. The in-phase (I) and quadrature (Q) parts of the signals are directly converted to 3 optical carriers by an optical IQ modulator. The PDM is emulated with a polarization controller (PC), a polarization beam splitter (PBS), a tunable optical delay line, and a polarization beam combiner (PBC). The delay of the optical delay line is exactly one OFDM symbol. Thus the 16-QAM PDM-OFDM superchannel has a total bandwidth of 37.5 GHz and a raw data rate of 225 Gb/s.
Receiver (Rx): Coherent detection is realized in the receiver (Rx). The LO is a tunable laser with a linewidth of 100 KHz. 4 balanced detectors (BD) are used to detect the polarization-diverse and phase-diverse signals output by the dual-polarization 90 degree hybrid. The electrical signals after the BDs are analog-to-digital converted by real-time oscilloscope (Tektronix 72004C) and then processed offline. The main digital signal processing procedures include digital low-pass filtering, frequency offset compensation, training-aided channel estimation, 2 × 2 multiple input multiple output (MIMO) channel equalization, and pilot-aided phase estimation.
OXC: We investigate the switching performance of OXCs with different filtering in three scenarios: the same superchannel passing through conventional OXCs, TOXCs with intra-WDM-grid spectrum switching, and cross-WDM-grid spectrum switching. Figure 3 shows the optical spectra and Q-factor performances passing through different OXC architectures. In scenario a (Fig. 3(a)), a superchannel centered at 1552.60 nm consisting of 3 subbands (Fig. 2-inset a) is generated and transported by all-SSS EON OXCs. The SSS is implemented with Finisar Waveshaper 4000S while the WSS is implemented with arrayed waveguide grating (AWG) with a channel spacing of 100 GHz. All the OXCs are demonstrated as 4 × 4 OXCs. In scenario b (Fig. 3(b)), the same superchannel is transported by CTS-TOXC and all the subbands are within one WDM grid. In scenario c (Fig. 3(c)), a superchannel is generated centered at 1552.95 nm consisting of 4 subbands (the third subband is a guardband for WDM-grids). The superchannel is transported by CTS-TOXC; it occupies the subbands in two spectrum-adjacent WDM-grids. The width of each subband is 12.5 GHz and the guardband in scenario b is also 12.5 GHz. The filtering profiles of SSSs and WSSs are shown as well.
Fig. 3 Optical spectra and Q factor: (a) a superchannel using EON-OXCs; (b) the same superchannel using CTS-TOXCs; (c) a superchannel using CTS-TOXCs and occupying two adjacent WDM-grids.
The back-to-back (BTB), after-one-OXC, and after-two-OXCs transmission performances are shown in Fig. 3. The sub-bands in scenario a switched by EON-OXCs suffer a Q-factor penalty about 0.2 ~0.4 dB per OXC (Fig. 3(a)). The sub-bands in scenario b and c switched by CTS-TOXCs except those adjacent to the guardbands of WDM-grid suffer a Q-factor penalty of 0.2 ~0.6 dB per OXC (Fig. 3(b) and 3(c)). These sub-bands suffer a little more penalty compared with EON-OXCs. Besides, the sub-bands in scenario c adjacent to the guardbands of WDM-grid suffer a Q-factor penalty of about 0.5 to 1.0 dB per OXC (Fig. 3(c)). This can be explained by the imperfect filtering characteristics of the WSSs (implemented by AWGs). The Q-factor degradation after passing 1 or 2 OXCs is different. This is because that we have used two different AWGs in our experiment and the wavelengths of these two AWGs are slightly inconsistent. After passing through two cascading OXCs, the Q-factors of all sub-bands are better than the forward error correction (FEC) limit (8.5 dB) [21], which validates the proposed TOXC architectures.
It should be noted that if we use 50 GHz AWG [22] or 50 GHz/100 GHz LCoS-based WSS [23] instead, the cutoff slope will be steeper and the spectrum slots at the edge of WDM grid will suffer slightly higher Q-factor penalty.
4. Spectrum allocation scheme for TOXC
In this section, we propose a heuristic spectrum allocation scheme with centralized path computation elements (PCEs) for on-line scenarios. The proposed contention-aware spectrum allocation (CASA) scheme can reduce the possible conflicts in WSSs caused by rigid WDM-grids and improve the blocking performance of TOXC. The network information about the SSS’s occupation situation in OXCs is stored in a centralized traffic engineering database (TED) and updated after every RSA process with periodically interchange. The CASA takes the new restrictions of TOXC into consideration and assembles the traffic demands with the same outgoing port to the same WDM grid by changing the selecting priority weight W.
The CASA scheme is described in Fig. 4(a) . In the conventional First-Fit scheme, the spectrum pattern with the minimized sequences number j will be selected. In contrast, the CASA scheme weights all the available spectrum patterns with parameter VPattern(j) to modify the selection priority. In CASA, all the available spectrum patterns along the routing path are weighted as:
where j is the sequence number of the first spectrum slot in the spectrum pattern and Wi(j) is the weight parameter representing the existing lightpaths’ direction information of the WDM-grid(s) occupied by the spectrum pattern in the WSSs of Node i:Examples of Wi(j) are shown in Fig. 4(b). Wi(j) = 0 means that the CASA scheme will not change the selection priority; Wi(j) = –w means that the CASA scheme will select the spectrum pattern (in the preferred WDM-grid) in advance of at most w spectrum slots; Wi(j) = ∞ means that the spectrum pattern cannot be used due to the WDM-grid occupation. The value of w will be detailed discussed in section V. The computational complexity of the proposed CASA scheme is O(c × h × J) (h is the routing hop number and J is the spectrum slot number per fiber).
It should be noted that we set a guardband of spectrum slot between all lightpaths. Moreover, if the spectrum pattern (consecutive spectrum slots for the incoming request including a guardband) crosses two adjacent WDM-grids, an extra guardband of the spectrum slot between the WDM-grids is set to cope with WSS filtering effect.
An example of VPattern calculation is shown in Fig. 5 : an incoming traffic which requires two spectrum slots is transmitted through one OXC. The current spectrum utilization is shown in Fig. 5(a) in which channel B has the same output port to the incoming request in this OXC. The weight of each spectrum pattern is defined in Fig. 5(b). The weights of j continuously increase as the spectrum slots’ sequence number increases; the amending parameter Wi(j) decreases when one of the WSSs has been occupied by a channel with the same direction (outgoing port) as the incoming request in the WDM-grid in this OXC, i.e. the second WDM-grid. The VPattern not only considers First-Fit principle which could use the spectrum in a more compact way, but also prefers allocating the channels with the same direction to the same WDM-grids to reduce the possible conflicts in WSSs for future traffics.
Fig. 5 An example of CASA scheme: (a) spectrum occupation situation; (b) spectrum patterns’ priorities.
5. Numerical simulations
We carry out extensive simulations to investigate the resource utilization efficiency of TOXC architectures. In EONs, bandwidth blocking probability (BBP) [24] is employed to evaluate the blocking performance. Considering the non-uniform bandwidth (or mixed line rate) traffic requests in EON, BBP is defined as below:
where Ci is the capacity of the demands with bandwidth i and Ai is its blocking probability. Furthermore, to evaluate the blocking performance more intuitively, we introduce the network accommodation (NA) [14], which is defined as the highest traffic load when BBP is lower than 0.1%.We investigate the BBP and NA of all-SSS EON OXCs, CTS-TOXCs, MS-TOXCs, as well as the WSS-based OXCs for reference. Numerical simulations are presented based on COST266 Pan-European network (26 nodes and 51 links). The fiber capacity is set with 400 spectrum slots. Similar to some previous studies [25,26 ], the generation of connection requests follows a Poisson process and the holding time for each connection follows a negative exponential distribution. The traffic bandwidth ranges from 1 to 10 spectrum slots and follows a uniform distribution. A spectrum slot is set as a guardband between every two adjacent pass-bands super channels. We assume that the connection requests are established with randomly selected source-destination node pairs. At least 106 connection requests are simulated in each scenario. The shortest path routing (SPR) scheme is used for all the scenarios. The First-Fit scheme is used for SSS-based OXCs. The CASA scheme is used for TOXCs. The WSS-based OXC is similar to CTS-TOXC (c = 1), and the spectrum allocation follows the First-Fit principle: the first available spectrum pattern with the same direction (outgoing port) will be allocated to the request.
We investigate the BBP and NA performance under different traffic loads with different value of c in Fig. 6(a) . We normalize all the BBP values so that they can be shown in the same figure. The simulation results show that c = 2 (when w = 0) is the most cost-effective choice. When c ≥ 3, TOXCs can slightly improve the BBP and NA performance but sharply increase the CAPEX, which linearly grows as c increases. Therefore, we choose 1 × 2 SSSs for cascading switching architectures in the rest of our simulation. The test of parameter w for the CASA scheme is shown in Fig. 6(b). When w = 0, the CASA is as same as First-Fit scheme. When w increases, the channels with the same direction will be more concentrated to the same WDM grid, but the effectiveness of First-Fit scheme and the concentration of channels in the entire spectrum will decrease. When w → ∞, the network spectrum fragmentation will be very serious and cause extra blocking. In different scenarios, a value of 50 is optimum for w according to our simulation. In different scenarios, w = 50 leads to the best performance according to our simulation. The comparison between proposed CASA scheme and the First-Fit scheme is shown in Fig. 7(a) . The CASA scheme is able to reduce the BBP, especially when traffic load is low (WDM OXC, relative 50% BBP improvement at 300 Erlang).
Fig. 7 Simulation results: (a) comparison between CASA and First-Fit. (b) BBP performances of different OXCs; (c) NAs of different OXCs; (d) transition cost of different OXCs.
In the simulation, the spectral grid of a SSS is 12.5 GHz (1 spectrum slot) and that of a WSS is 100 GHz (8 spectrum slots). The BBP performance of OXCs is shown in Fig. 7(b). The guardbands between WSSs’ spectral grids cause some extra BBP, but our TOXCs still have significant advantages over WDM OXCs. As shown in Fig. 7(c), we normalize all the NA results by that of the SSS-based OXCs. The proposed CTS-TOXCs can provide about 86.2% NA compared with all-SSS EON OXCs, while MS-TOXCs can provide about 91.3% NA.
The CAPEX of the network includes the fiber deployment and the CAPEX of the OXC. Fibers in the WDM network should be reused to avoid the high cost of additional fiber deployment. Therefore, the CAPEX we consider is indeed the CAPEX of the OXC, and there is a tradeoff between high NA and low CAPEX of the OXC. The CAPEX of OXCs in COST266 Network is calculated in Eq. (4) based on the OXC degrees in Table 1 . ni represents the node degree in node i. As the large-port-count SSSs are cascaded by small ones, we normalized the CAPEX of OXC by its overall node port count. Compared to the all-SSS EON OXC, the proposed TOXC can save SSSs’ port count especially when the OXC port count is large. In the COST266 topology, the proposed TOXC can save the SSSs’ port count by 56.4% (CTS-TOXC) and 28.2% (MS-TOXC). As shown in some previous studies (e.g [14].), the highest port count commercially available at present is 20 + and will be even higher in the future. When n = 20, the proposed TOXCs can save the SSSs’ port count by 90% (CTS-TOXC) and 45% (MS-TOXC) compared with all-SSS EON OXCs.
Table 1. OXC Degrees in COST266 Network
Furthermore, we evaluate the transition cost of the entire OXCs (including the WDM OXCs): Compared with LCoS-based 10 GHz SSS, 100 GHz WSS is based on some more cost-effective technologies, e.g. planar lightwave circuit (PLC). For clarity, we normalize the cost by WSS’s port and assume the cost ratio γ of SSS to WSS from 1 to 3. The normalized CAPEX of the n × n OXCs is shown in Eq. (5).
The transition cost of OXCs is shown in Eq. (6). For TOXCs, because of the reused WDM-OXC, the CAPEX of TOXCs is reduced by the CAPEX of WDM-OXCs. The evaluation of transition cost of OXCs is shown in Fig. 7(d). As we can see, when the OXC port count is higher than 6, the proposed TOXCs always have a CAPEX advantage comparing with all-SSS EON OXCs; this advantage is more significant when cost ratio of SSS to WSS is higher. When the OXC port count is 20 and the cost ratio γ is 3, the TOXCs can save the transition cost by more than 50%.
6. Conclusion
In this paper, we propose and experimentally demonstrate a transition OXC (TOXC) structure for EONs by cascading SSSs with WSSs. We evaluate the blocking performance of proposed TOXCs in different scenarios by simulation. The proposed TOXC architectures can significantly save SSSs’ port count n while only slightly decreasing BBP performance and network accommodation compared with the all-SSS EON OXC architectures. When n = 20, the proposed TOXCs can save the SSSs’ port count by 45% (CTS-TOXC) and 90% (MS-TOXC), and the transition cost by about 50%.
Acknowledgment
This work was supported by National Basic Research Program of China (973 Program, No. 2014CB340105 and No. 2012CB315606) and National Natural Science Foundation (NSFC) (No. 61377072, No. 61275071 and No. 61205058).
References and links
1. W. Shieh, X. Yi, and Y. Tang, “Transmission experiment of multi-Gi-Gabit coherent optical OFDM systems over 1000 km SSMF fiber,” Electron. Lett. 43(3), 183–185 (2007). [CrossRef]
2. A. Lowery, L. B. Du, and J. Armstrong, “Performance of optical OFDM in ultralong-haul WDM lightwave systems,” J. Lightwave Technol. 25(1), 131–138 (2007). [CrossRef]
3. J. Armstrong, “OFDM for optical communications,” J. Lightwave Technol. 27(3), 189–204 (2009). [CrossRef]
4. M. F. Huang, A. Tanaka, E. Ip, Y. K. Huang, D. Qian, Y. Zhang, S. Zhang, P. N. Ji, I. B. Djordjevic, T. Wang, Y. Aono, S. Murakami, T. Tajima, T. J. Xia, and G. A. Wellbrock, “Terabit/s Nyquist superchannels in high capacity fiber field trials using DP-16QAM and DP-8QAM modulation formats,” J. Lightwave Technol. 32(4), 776–782 (2014). [CrossRef]
5. G. Bosco, V. Curri, A. Carena, P. Poggiolini, and F. Forghieri, “On the performance of Nyquist-WDM terabit superchannels based on PM-BPSK, PM-QPSK, PM-8QAM or PM-16QAM subcarriers,” J. Lightwave Technol. 29(1), 53–61 (2011). [CrossRef]
6. B. Kozicki, H. Takara, T. Yoshimatsu, K. Yonenaga, and M. Jinno, “Filtering characteristics of highly-spectrum efficient spectrum-slicedelastic optical path (SLICE) network,” Proc. OFC2009, JWA43.
7. M. Jinno, H. Takara, B. Kozicki, Y. Tsukishima, Y. Sone, and S. Matsuoka, “Spectrum-efficient and scalable elastic optical path network: architecture, benefits, and enabling technologies,” IEEE Commun. Mag. 47(11), 66–73 (2009). [CrossRef]
8. H. Takahashi, K. Takeshima, I. Morita, and H. Tanaka, “400-Gbit/s optical OFDM transmission over 80 km in 50-GHz frequency grid,” Proc. ECOC2010, Tu.3.C.1. [CrossRef]
9. Z. Chen, Y. Chen, J. Li, P. Zhu, Y. Xu, B. Guo, and Y. He, “Demonstration of elastic optical networks based on OFDM/SCFDM subbands,” Proc. ACP2013, AF1H.3. [CrossRef]
10. http://www.finisar.com/products/optical-instrumentation/WaveShaper-4000S
11. A. A. M. Saleh and J. M. Simmons, “Technology and architecture to enable the explosive growth of the Internet,” IEEE Commun. Mag. 49(1), 126–132 (2011). [CrossRef]
12. P. Pagnan and M. Schiano, “A λ switched photonic network for the new transport backbone of Telecom Italia,” Proc. PS2009, ThII2–1. [CrossRef]
13. S. Woodward, “Balancing costs & benefits in a flexible grid network,” in Proc. OFC/NFOEC (IEEE, 2012), paper WS.
14. Y. Iwai, H. Hasegawa, and K. Sato, “A large-scale photonic node architecture that utilizes interconnected OXC subsystems,” Opt. Express 21(1), 478–487 (2013). [CrossRef] [PubMed]
15. T. Ban, H. Hasegawa, K. Sato, T. Watanabe, and H. Takahashi, “A novel large-scale OXC architecture that employs wavelength path switching and fiber selection,” in Proc. ECOC (IEEE, 2012), paper We.3.D.1.
16. L. Hai-Chau, H. Hasegawa, and K. Sato, “A large capacity optical cross-connect architecture exploiting multi-granular optical path routing,” in 2012 International Conference on Photonics in Switching (IEEE, 2012), FrS26.
17. K. Sato and H. Hasegawa, “Optical networking technologies that will create future bandwidth abundant networks,” J. Opt. Commun. Netw. 1(2), A81–A93 (2009). [CrossRef]
18. Y. Yamada, H. Hasegawa, and K. Sato, “Coarse granular routing in optical path networks and impact of supplemental intermediate grooming,” Proc. ECOC 2010, Th.10.F.1. [CrossRef]
19. Y. Taniguchi, Y. Yamada, H. Hasegawa, and K. Sato, “A novel optical networking scheme utilizing coarse granular optical routing and fine granular add/drop,” in Proc. OFC/NFOEC (OSA, 2012), JW2A.2. [CrossRef]
20. S. Mitsui, H. Hasegawa, and K. Sato, “Low loss and cost-effective hierarchical optical path cross-connect switch architecture based on WSS/WBSS,” in 2009 International Conference on Photonics in Switching (IEEE, 2009), paper FrII2–3. [CrossRef]
21. W. Peng, T. Tsuritani, and I. Morita, “Transmission of high-baud PDM-64QAM signals,” J. Lightwave Technol. 31(13), 2146–2162 (2013). [CrossRef]
22. http://www.ntt-electronics.com/en/products/photonics/awg_mul_d.html
24. X. Wan, N. Hua, and X. Zheng, “Dynamic routing and spectrum assignment in spectrum flexible transparent optical networks,” J. Opt. Commun. Netw. 4(8), 603 (2012). [CrossRef]
25. H. Hasegawa, Y. Tanaka, K.-I. Sato, and Y. Iwai, “Subsystem modular OXC architecture that achieves disruption free port count expansion,” Proc. ECOC2013, Th.2.E.4. [CrossRef]
26. A. Fallahpour, H. Beyranvand, S. A. Nezamalhosseini, and J. A. Salehi, “Energy efficient routing and spectrum assignment with regenerator placement in elastic optical networks,” J. Lightwave Technol. 32(10), 2019–2027 (2014). [CrossRef]
