1
Evolution of dynamic optical networks Naoya Wada and Hideaki Furukawa Photonic Network Research Institute, National Institute of Information and Communications Technology (NICT), 4-2-1 Nukui-kitamachi, Koganei-shi, Tokyo, 184-8795 Japan. E-mail:
[email protected]
Abstract—In this paper, we show the recent progress of physical layer technologies for the dynamic optical networks. There is much activity in the design of post-Internet or future networks. These are based on new design concepts that look beyond the next generation network (NGN) and the Internet. The future network will maintain the sustainability of our prosperous civilization and help resolve various social issues and problems by the use of information and communication technologies (ICT). In order to realize the future network, many novel technologies in the physical layer are required, in addition to technologies in the network control layer. In the physical layer, it is very important that especially technology to support real dynamic network operation, such as grid free, format free, and burst mode in both time and spectrum domain. An optical packet and circuit integrated network system is demonstrated as an examples of cutting-edge network system including many new physical layer technologies for the dynamic operation. Other related activities of dynamic optical networks will be also shown in the presentation. Index Terms—Future networks, dynamic optical networks, physical layer technologies, optical packet and circuit integrated network.
I. INTRODUCTION The remarkable advances in telecommunications technology in recent years have brought about a new information revolution that ranks alongside the industrial revolution. Today, the Internet is an essential part of our social infrastructure not only in the business world but also in our everyday lives. The Internet, however, is facing a crisis. The Internet was invented in the late 1960s, originally for use as a kind of communication tool for closed research communities and for communication links among computers. Today, around 1,000,000,000 hosts are connected to the Internet, and the number is still increasing. The Internet architecture has been extended in an attempt to accommodate boundless user demand, but the thin veneer of expansion of the Internet has come off, and it is now difficult for the Internet to respond to newly emerging social demands. There is much activity in the design of post-Internet or future networks such as European Future Internet Forum [1], Exploring networks of the future (GENI) [2], EC ICT
Research in FP7 [3], National Science Foundation (NSF) NeTS Future Internet Design (FIND) Initiative [4], and NeW Generation Network (NWGN) [5] around the globe. These are based on new design concepts that look beyond the next generation network (NGN) and the Internet. These future networks will maintain the sustainability of our prosperous civilization and help resolve various social issues and problems by the use of information and communication technologies (ICT). In short, it aims to fundamentally solve the difficult issues and limits in an improved and extended Internet by taking a clean-slate design approach, unconstrained by existing technologies. For example, research and development of the NWGN is now being conducted as a Japanese national project [5]. As for the medium- to long-term research and development (R&D) strategy, the National Institute of Information and Communications Technology (NICT) published its NWGN vision, involving five network targets. The vision is derived from discussions about what and how NICT should contribute to create a prosperous society of the future. The five network targets represent functional network requirements. They are “Value Creation Network,” “Trustable Network,” “Ambient/Ubiquitous Network,” “Self-configuring Network (Connectable network without restrictions),” and “Sustainable Network.” It should be noted here that our technological strategy is not seeds-oriented but rather needs-oriented. General technological strategies and roadmaps have been established using an incremental approach that is an extension of today’s technology [6]. AKARI is an NWGN architecture project. AKARI has produced designs of several enabling components, such as optical packet and circuit integrated networking for diversity inclusion, network virtualization for sustainable networking, a regional wireless/sensor platform network for reality connection an identifier (ID)/locator split network architecture, and robust and self-organized networking [7]. In order to realize the future network, many novel technologies in the physical layer are required, in addition to technologies in the network control layer. In the physical layer, it is very important that especially technology to support real dynamic network operation, such as grid free, format free, and burst mode in both time and spectrum domain. In this paper, we show the recent progress of physical layer technologies for the dynamic optical networks. An optical packet and circuit integrated network system is demonstrated as an examples of cutting-edge network system
2 including many new physical layer technologies for the dynamic operation. Other related activities of dynamic optical networks will be also shown in the presentation. II. PHOTONIC TECHNOLOGIES FOR PHYSICAL LAYER IN THE FUTURE NETWORK
In the future network, high scalability, fine granularity, and flexibility will be required, in addition to increased network capacity [7, 8]. In order to satisfy these demand, it is very important that physical layer technology to support real dynamic network operation, such as grid free, format free, and burst mode in both time and spectrum domain. Optical channels of varying granularity, such as path/burst/packet in the time domain and narrow/wide/bundle in the spectral domain, should be used as a pool of physical layer resources depending on the application [7]. The basis of the physical layer of a photonic network is transmission technology in the links and switching technology in the nodes. The transmission capacity per fiber can be increased to over 100 Tbit/s by using various optical multiplexing methods and modulation formats, such as differential quadrature phase shift keying (DQPSK) and quadrature amplitude modulation (QAM) [9, 10]. In addition, orthogonal frequency division multiplexing (OFDM) technology [11] has the potential to enable flexible-bandwidth transmission. This concept with flexible-bandwidth network operation is called Elastic Optical Network [11] and become a very popular research subject. A more general and versatile solution for implementing flexible bandwidth capable transmitters and receivers has been demonstrated based on dynamic optical arbitrary waveform generation (OAWG) [12] and measurement (OAWM) [13] technologies. Recently, a flexible bandwidth network system using OAWG and wavelength selective switch (WSS) [14] and a flexible bandwidth, grid-less, and modulation formats free optical switching system [15] have been also demonstrated. All these demonstrations are technologies for optical path networks. On the other hand, it is more difficult to provide these flexible function and dynamic operation in the packet based network because of complicated processing in the network node. The node-throughput is generally limited due to electronic processing in the node systems. In current electronic high-end IP routes, layer-1 and layer-2 switches, high-speed optical signals are received and de-multiplexed into low-speed electrical signals and parallel electronic processing is employed. However, many line cards and a large-scale electronic switch causes a serious power consumption problem. In addition, as the modulation formats of optical signals become more sophisticated, more complex receivers become necessary. An optical packet switching (OPS) system, which can implement high-throughput forwarding of optical packets without optical-to-electrical-to-optical (O/E/O) converters in the physical layer, is quite attractive for node systems with energy-efficient processing and transparency for various bit rates and formats. Despite the limited functionality of optical technologies, various OPS systems have been developed for more than 15 years to exploit the potential abilities of OPS
systems [16]–[28]. Although OPS systems themselves have not been used in commercial networks at present, they play an important role in the research activity as a vehicle leading to many novel basic technologies. These novel technologies are necessary to realize the physical layer of the dynamic optical network system, as well as OPS systems. There has also been significant evolution in the hardware systems to support real dynamic network operation, such as grid free, format free, and burst mode in both time and spectrum domain. Many important functions, which were difficult to achieve 10 years ago, have been developed and are now used in prototype-level demonstrations. III. AN OPTICAL PACKET AND CIRCUIT INTEGRATED NETWORK A. Problems in dynamic operation In order to provide diversified services such as best-effort and quality of service (QoS) guaranteed services, we have proposed an optical packet and circuit integrated network based on common wavelength and fibers [7]. OPS provides bandwidth-sharing and best-effort data transferring, and optical circuit switching (OCS) provides an occupied bandwidth and end-to-end QoS guaranteed data transmission [29-30]. By dynamically sharing wavelength resources for OPS or OCS links [31], new or urgent services are supported. By multiplexing control packets for signalling and resource control on OPS links, extra interfaces are decreased and networks are simplified. In a proposed integrated network, optical packets and data on lightpaths are multiplexed and also transmitted on the same infrastructures as shown in Fig. 1.
Fig. 1. Concept of an optical packet and circuit integrated network.
We have demonstrated an OPS/OCS node system [29-30] based on OPS system [27, 28] and signalling and routing system on control plane for OCS networks [31]. In integrated networks, its dynamic operation leads to problems in the physical layer. For example, power fluctuation occurs because the dynamic change of the number of lightpaths and the packet traffic causes some transient gain excursion of erbium-doped fiber amplifiers (EDFAs) set on transmission lines and systems. To reduce this effect, we have developed an all-bandwidth burst-mode amplifier system and a high dynamic-range packet receiver for the tolerance of the packet power fluctuation. These technologies are very useful for any dynamic optical network, as well as integrated network.
3
Fig. 2. Amplified optical packets in dynamic lightpath setup/release..
B. Key technologies for tolerance of power fluctuation In edge nodes of an integrated network, low-band signals such as Ethernet frames are en/decapsulated with wide-band optical packets. We use a colored optical packet, which consists of 10 Gbit/s optical signals of N wavelengths. Also, control optical packets for path signalling and resource control are exchanged via OPS links. End-to-end lightpaths are provided for high-quality services on user demands. Wavelength resources are used as occupied resources of OPS and OCS links, and shared resources. The shared resources are allocated to OPS or OCS links depending on demands for lightpath or packet transferring. Due to the dynamic lightpath setup/release, the change of packet traffic, and the dynamic allocation of shared resources, the total input power to an EDFA is changed and the transient gain excursion occurs. Then, the power of amplified packets or lightpaths is fluctuated. Figures 2 and 3 show examples of the packet power fluctuation. We input 80 (8λ x 10) Gbit/s colored optical packets and 8-wavelength lightpaths into an EDFA, and extracted only optical packets by a band-pass filter after amplification. Figures 2(a) and 2(b) show the spectrum of input 80 Gbit/s optical packets and 8- lightpaths, and the temporal waveform of extracted packets, respectively. Then, we released 6-lightpaths. Figures 2(c) and 2(d) also show the spectrum of input packets and 2-lightpaths, and the temporal waveform of extracted packets, respectively. The packet power increased by about 1 dB. We can confirm that the packet power changes due to the dynamic lightpath setup/release. Next, we changed only the duty cycle of a packet-sequence as shown in Fig.3(a). From results in Fig.3(b), the packet power also changes depending on the duty-cycle. These power fluctuations may cause bit-errors in receivers. To tolerate the power fluctuation, we developed a high dynamic range packet receiver (HDR-PR), which consists of a PIN-PD, a transimpedance amplifier for sensitive detection, and a burst-mode limiting amplifier with variable thresholding (Fig.4(a)). While a previous Rx. using OPS prototype has 5.1 dB dynamic range [26, 27], the new one has 14.1 dB range. Figure 4(b) shows a recovered packet by the HDR-PR from a power-fluctuated packet. We adopted a single phase lock loop configuration to enable instantaneous synchronization of the output clock phase because packets are bursty input. On the contrary, to eliminate optical surge and gain transient for shorter packet, (~100ns) we developed transient-suppressed (TS-) EDFA [32]. However, long term transient still remained.
Fig. 3. Amplified optical packets in changing packet duty cycle.
To uniformly amplify streaming packets, we developed all-bandwidth burst-mode amplifier system based on TS-EDFA with external controlling techniques such as optical feed back loop, automatic gain control (AGC) and automatic level control (ALC) (Fig.5(a)). Also, impairment came from multi-wavelength add/drop can be mitigated effectively [33]. Figure 5(b) shows almost uniformly amplified packets by a developed amplifier compared with results in Fig.3(b) at various duty cycles. C. Demonstration of optical packet and circuit simultaneous transmission Figure 6(a) shows the demonstration system, which consists of an integrated OPS/OCS node as a core node, two edge nodes, and some clients which send data by IP-packets or lightpaths, including a camera and a display for uncompressed HDTV transmission. The TS-EDFAs are used to amplify coupled optical packets and lightpaths in fiber lines. To uniformly amplify optical packets, TS-EDFAs with AGC/ALC are used only for OPS links after dividing packets and lightpaths in each node. In addition, HDR-PRs are used to have a tolerance for packet power fluctuation. The shared resource, the occupied resources for OPS and OCS are 1538.9-1541.3 nm (λ9-λ12), 1547.7-1553.3 nm (λ1-λ8), and 1558.9-1561.4 nm (λ13-λ16), respectively. An IP-packet Tx. sends IP-packets with the destination IP address of the core node for OCS control and ones with the destination IP address of an IP-packet Rx. for data. 8-Lightpath Tx./Rx. generate and receive 1.25 Gbit/s signals of 8-wavelengthes on data-plane. Here, OCS control signals from a Lightpath Tx./Rx. are sent via an IP-packet Tx.. In Edge node #1, these IP-packets are encapsulated into 80 Gbit/s colored optical packets with 1.24 Gbit/s optical labels corresponding to the IP address by an IP-OP converter [34]. Then, these packets and 8 lightpaths are coupled and launched into a field fiber line. The field fiber lines are located between Koganei and Otemachi in Tokyo with loop-back configuration at JGN2plus (present JGN-X) [35]. One round-trip fiber line consists of 85 km field installed single-mode fiber (SMF) with 27 dB loss and 25 km dispersion compensating fiber (DCF). The dispersion was compensated at -20 ~ +25 ps in C-band as shown in Fig. 6(b). Figure 6(c) shows 2.5 ps or less fluctuation of differential-group-delay (DGD) in one round-trip field fiber measured at 10 times in 1 hour. As a reference, the DGD fluctuation of 100 km our bobbin fiber was about 0.2 ps as shown in Fig. 6(d). This DGD fluctuation causes skew,
4
Fig. 4.High dynamic-range burst-mode packet receiver.
Fig. 5. Transient-suppressed EDFA with automatic level control.
Fig. 6. Demonstration system and characteristics of field installed fiber.
polarization mode dispersion, and other effects on colored optical packets. These effects are in a margin of OP-IP converters at core node system and Edge node #2. If the DGD is get larger, an OP-IP converter has no problem because of its input electrical buffer, but optical switches may need more guard time or additional compensation. In the OPS/OCS node, whose the detail of configuration is shown in Ref.[29, 30], two 2 x 2 OPSs and two 2 x 2 OCSs are used for the occupied and the shared resources, respectively. Optical packets and 8 lightpaths are divided by a waveband selector (WBS). A resource controller makes 8x8 switches (resource switches) connect the optical signals with OPS or OCS, respectively. Control packets with the destination address of core node are forwarded to this node by a label processor and a 1 x 2 LiNbO3 switch (LN-SW) in OPS. Then, a lightpath setup starts, or the shared resource is re-allocated. An arrayed waveguide grating (AWG) in the OCS divides the waveband into 8 lightpaths, and the large-scale switch forwards each signal to an output port. The
core node also sends IP-packets to Edge node #2 for OCS control. Control packets from the core node and data packets from Edge node #1 are input into a 2 x 1 optical buffer which consists of plural LN-SWs and fiber-delay-lines. To avoid packet collisions between their packets, the buffer adequately delayed packets. Figure 7(a) shows packet sequences in above-mentioned operations at the input/output port, and the optical buffer of the core node. We confirmed that only control packets with the destination label of the core node were switched to the core node and that the buffer operated normally. Figure 7(b) shows the spectrum at the output of the core node. Figure 7(c) shows the temporal waveform of optical packets and lightpaths after 170 km transmission. In Edge node #2, HDR-PRs receive both data and control optical packets. Then, an OP-IP converter extracts IP-packets from optical packets and sends to the IP-packet Rx.. The IP-PER for data transferred from IP-packet Tx. was 3.33 x 10-5 as shown in Fig.7(d), which is high quality transmission. Through ICMP sending/receiving (i.e. ping) and 1.6 Gbit/s HDTV
Fig. 7. Experimental results.
5 transmission on 8-lightpaths, we confirmed that OCS control packets were safely reached to all nodes and 8 lightpaths were successfully set up. We developed a high dynamic range packet receiver and a TS-EDFA with automatic level control for dynamic operation of optical packet and circuit integrated network system. Field transmission and switching of 80 Gbit/s colored optical packets and 8-lightpaths has been demonstrated [33]. IV. AN OPTICAL PACKET AND CIRCUIT INTEGRATED RING NETWORK TESTBED
We have developed an optical packet and circuit integrated ring network testbed with polarization-independent 4×4 optical switches and gain-controlled optical amplifiers [36]. Figure 8 shows a novel integrated OPS/OCS node, which consists of mainly six devices: 10G-OTN transponders, a 100G-OP transponder, two wavelength selective switches (WSS), an 4 × 4 SOA switch subsystem, a switch controller and optical amplifiers. In OCS links, a 10G-OTN transponder encapsulates 10GbE frame from client side into OTN format. On the other hand, in OPS links, we introduce 100 Gbps colored optical packets, consists of multiple 10 Gbps optical payloads of different 10-wavelengths as shown at the bottom of Fig.9. In a 100G-OP transponder, an incoming 10GbE frame from client side is encapsulated into ten 10 Gbps optical payloads of an optical packet as shown in Fig.9. The packet length is variable corresponding to the frame length. In parallel, a 10Gbps 8 Byte route header including an 8 bit destination Node-ID of the optical packet is attached in one payload. The destination Node-ID is determined according to a mapping table between destination Node-IDs and the destination IP address of incoming frames. The bit-pattern matching of IP address within the arbitrary range is possible by mask and offset functions. In an integrated network, wavelength resources are divided by waveband, and each of these wavebands is allocated to OPS or OCS links. In an integrated OPS/OCS node, two WSSs are used for combining/dividing of OPS and OCS wavebands. In addition, for OCS links, two WSS work as add/drop multiplexers. The WSSs are controlled to adequately set up optical paths. For optical packet forwarding, a switch controller reads the destination Node-ID of the route header and controls a 4 × 4 SOA switch subsystem to forward an
Fig.8. Integrated Optical packet and circuit switch node.
optical packet to an adequate output port according to a switching table of each input port. The 4 × 4 SOA switch subsystem with a broadcast-and-select configuration is developed. This switch not only has several nano-second switching-speed, low polarization-dependency and loss compensation, but also handles 100 GHz channel spacing colored optical packets without the crosstalk caused by a four wave mixing effect. We measured the spectrum of ten optical payloads of a forwarded 100 Gbps optical packet at output of an OPS/OCS node. Figure 10 shows the maximum and minimum peak-intensity in case of randomly rotating the polarization of the optical packet by a polarization controller in input of the OPS/OCS node. About 2 dB intensity-difference is the polarization-dependency of the OPS/OCS node including one SOA switch subsystem. To eliminate optical surge and gain transient for shorter packet, (~100ns) we have developed a transient-suppressed Erbium doped fiber amplifier (TS-EDFA) with enhanced active erbium area [33]. In addition, we improve the TS-EDFAs by optimizing the fiber length, the core diameter and the Erbium density of EDF and adjusting the power of a pump laser-diode to suppress the wavelength-dependent gain characteristic and the output power fluctuation due to changing packet-rate. We have demonstrated error-free (frame-error-rate
