You are on page 1of 14

Theoretical and Applied Informatics ISSN 1896–5334 Vol.21 (2009), no. 3-4 pp.

167–180

Overview of optical packet switching
a ´ I RENEUSZ S ZCZE SNIAK

IITiS Politechnika Cz˛ estochowska ul. Dabrowskiego ˛ 73 42-200 Cz˛ estochowa
Received 15 October 2009, Revised 30 October 2009, Accepted 5 November 2009

a

Abstract: Optical packet switching has been researched for about two decades, but it has not been deployed in commercial networks yet. Nonetheless, the research on optical packet switching continues as it promises to perform better than electronic hardware. We report on the established results of optical packet switching, and examine the reasons for its current state. Keywords: optical networks, optical packet switches, optical buffers

1. Introduction Communication networks are necessary in modern societies. Their role is to relay information quickly and reliably. A communication network is built of nodes that process information, and links between nodes, that transport the information. Communication networks can be broadly divided into radio networks, electrical networks, and optical networks. The backbone communication networks are usually optical. The volume of data transported by optical networks is growing, the service level agreements (SLA) are getting stricter, while the pressure on lowering costs keeps strong. Therefore the optical networks are constantly being expanded and improved, and new hardware solutions and control protocols are being developed. The current optical networks use only a small fraction of the bandwidth offered by optical fibers, and one of the reasons for it are slow electronic components. So far, the replacement of electronic components with their optical counterparts has been beneficial. An excellent book on optical networks and their evolution is [32], which describes in a clear and practical manner a wide range of topics, starting from optical physical laws to the management of optical networks. Another book on optical communication is [27], which concentrates on the recent research results.

The introduction of the wavelength division multiplexing (WDM) was groundbreaking. SONET/SDH or IP. Additionally. The WDM networks are build mainly of OXCs and OADMs. which can add or drop a single wavelength from a fiber. which enabled optical amplification of many wavelengths simultaneously. While WDM resulted in more bandwidth. At nodes the optical signal was converted to the electrical signal. but now of many wavelengths. then the data already received is buffered at the node. The optical transmission employed one wavelength only. which in turn reduced of the number of OEO regenerators required. and sharing it with the nodes through which the lightpath transverses. This conversion is called the optical-electrical-optical conversion OEO. A number of solutions have been proposed that are capable of switching below a wavelength [5]: • It has been proposed to establish a lightpath between two network nodes. • In optical burst switching it was proposed to establish lightpaths for a short period of time. The optical reach increased as the technology advanced. even for a few milliseconds. WDM networks are able to establish lightpaths between network nodes. but in the optical cross connects (OXC) it can be switched optically. The lightpath does not have to be entirely established for the sender to start transmission. because it enabled the transmission of a number of wavelengths in a single fiber. it can request the network to tear down the lightpath even if the destination node has not received all the data. When the source node has finished transmitting. between many source nodes and a single destination node called time-domain wavelength interleaved networking [37]. still the expensive OEO regeneration was required. optical fibers were used for short distance communication. In [4] the history of optical networks is described. and then converted back to the optical signal to be sent to the next node. A disadvantage . If at some node the lightpath has not been switched. there are available optical add-drop multiplexers (OADM). and so the required switching times at nodes equal a few milliseconds. which are trailed to the networks at the higher layer: PDH. 6]. An optical burst is longer than an optical packet. In regular electric switches a wavelength has to be converted to the electric signal. A number of solutions have been proposed for establishing unidirectional connections between a single source node and many destination nodes called super lightpaths [25]. and their evolution towards supporting electronic client networks. with the support of optical amplifiers. and between many source nodes and many destination nodes called optical light-trails or distributed aggregation [16. If the distance between nodes exceeded the optical reach. and not microseconds as in packet switching. The introduction of optical amplifiers was another ground-breaking event.168 At the beginning of the eighties of the twentieth century. regenerators and wavelength converters. then OEO regenerators were deployed.

OPS hardware could lower power requirements. STOLAS [26]. • Another proposed solution is optical packet switching (OPS) that we describe in the section that follows. and sent to the access node. without the OEO conversion. 39. most notably KEOPS [12]. Next the packets are converted to the optical signal and sent to the core network. In an access node clients of the network operator receive service. dissipate less heat and take less space compared to electronic equipment [31]. or in other words. Optical packet switching In optical packet switching the payload (the user data) is switched optically. In the destination edge node. whose hallmark is switching many packets on different wavelengths simultaneously [46]. which in turn are connected to core nodes (CN). 2. . In an edge node the incoming packets (mainly IP packets. We can expect OPS to eventually replace traditional electronic switching. OPS can be faster. Moreover. and others. Another version of OPS is photonic slot routing. Davis [45]. The overview of switch architectures is presented in [36. The overview of hardware techniques needed in OPS is discussed in [3]. they make hops between nodes. 44]. i.e. There has been a number of large projects carried out on OPS. [22] are grouped into optical packets for better utilization of the optical core and for lowering the disadvantageous effects of self-similarity [23]. If the packet header is processed optically. The overview of the early works on OPS is discussed in [21]. However. If the packet header is processed electronically. which highers the requirements for OPS. then this version of OPS is called all-optical packet switching. Electronic access nodes (AN) are connected to edge nodes (EN).169 of burst switching is the necessity of maintaining a low network utilization level in order to guarantee a satisfactory service availability. and also cheaper to purchase and maintain than traditional switching with the OEO conversion [12]. the wellestablished SONET/SDH technology is being improved to transport the data traffic better [14]. then this version of OPS is called transparent OPS. Even though OPS progressed considerably. electronic equipment is getting cheaper and more efficient. 24. DAVID [10]. 1. the packet is converted to the electric signal. still this technology is unlikely to be deployed in a few years time. projects at the University of California. is more reliable and consume less energy [31] in comparison with their electronic counterparts. Packets are sent between core nodes. because optical network equipment is cheaper to maintain. A general model of an optical packet-switched network is shown in Fig. Core nodes switch packets optically.

The existing optical gates can be switched on and off in about two nanoseconds. several works investigated practical applications of the current OPS technologies [13. There are companies that offer optical packet switches. The duration of a time slot depends on the technology used in the equipment. and the duration of a packet is not longer than the time slot. Nonetheless. and how OPS could transport IP traffic [17. the time it takes to change the switch state from “=” to “x” or from “x” to “=”. In [31] this technology is criticized for having failed to meet practical expectations. The Yokogawa Electric Corporation offers an optical packet switch with two ingress fibers and two egress fibers.e. . The switching time. “EN” is an edge node. 1. 20]. 15. This technology received less attention than the synchronous technology. equals about two nanoseconds. 35. 22. In [11] a similar criticism is delivered. which corresponds to the guard time between packets. where “AN” is an access node. OPS can be either synchronous or asynchronous: • In synchronous OPS the time domain is divided into slots during which packets are sent and received. in the KEOPS project the time slot lasts 1. 30]. 28]. where each fiber carries one wavelength [43]. • In asynchronous OPS the time domain is slotted.170 AN AN EN AN CN CN CN AN EN core network CN AN EN AN AN AN AN Fig. 8. 41. For instance.64µs [12]. i. and so packets can arrive at any time. Optical packet-switched network. and “CN” is a core node A practical perspective on OPS is offered in a number of papers. Some practically oriented works on optical networks even do not mention optical packet switching [19. The switch is used to build rings in the core. but that still promises to deliver in the future.

as there are wavelengths available in the output fiber. and so that the clients receive the required quality of service. 2 shows the symbols of the elements. so that the core network is not overloaded. an optical amplifier. If there are more packets requesting this output fiber. because the asynchronous OPS suffers of larger contention probabilities in comparison with the synchronous technology [10]. Access control and routing algorithms play an important role in OPS. The output fiber carries several wavelengths. which copes with the lack or scarcity of optical buffers by sending a packet to an output fiber that is not requested by it. In a time slot a switch can send to an output fiber up to as many packets. a splitter. one wavelength per each output fiber. Deflection routing can improve network performance for light loads [40]. where each of the wavelengths is taken selectively from the input fibers. The other reason might be the low network utilization needed. The routing algorithm is responsible for forwarding packets to the appropriate output fibers on the appropriate wavelength.171 even though it is easier to build. a fixed wavelength converter. Components he basic optical components of OPS are a multiplexer. A number of contention resolution algorithms have been proposed. 2b) has the opposite function to a multiplexer. A demultiplexer (Fig. so that the network delivers packets with low probability of packet loss. A routing algorithm directs packets toward their destination nodes. Light travels through these elements only in one direction. a coupler. One of the methods for resolving contention among packets is deflection routing. a 2 × 2 switch. then the switch needs to resolve the contention among packets. . but it can also worsen network performance for heavy loads. an optical gate. 2a) has several input fibers and a single output fiber. The action of sending a packet to a wrong output fiber is called a deflection. an optical delay line. because no synchronization is needed [11]. Wavelengths from the input fiber are sent to the output fibers. because then packets can be deflected many times causing a livelock [7]. a tunable wavelength converter. Deflection routing prevents packets from being dropped by using available output fibers as buffers. a demultiplexer. Fig. and an N × N arrayed waveguide grating (AWG). It has probably been so. A multiplexer (Fig. A multiplexer has a single input fiber and several output fibers. and tries to optimize the use of the scarce OPS resources (such as buffers and wavelength converters). 3. From each input fiber a single wavelength is selected and sent to the output fiber. An access control algorithm is responsible for managing packets that request admission to the core network from the access network. which are usually the DiffServ rules for establishing the order of sending the packets [2].

(c) coupler. 2. 2d) has the opposite function to a coupler. The light sent to the output fiber is of a fixed wavelength. . Fig. A 2 × 2 switch (Fig. A tunable wavelength converter (TWC. The switch is controlled electrically in a few nanoseconds. In comparison to the SOA. (i) optical delay line. 2h) amplifies the incoming optical signal in the given wavelength range. (h) optical amplifier. A splitter (Fig. An optical gate (Fig. (f) tunable wavelength converter. A fixed wavelength converter (FWC. 2f) is capable of tuning the wavelength at the output fiber. (g) optical gate. 2c) has several input fibers and a single output fiber. Fig. (b) demultiplexer. unfortunately. 2g) amplifies or damps the incoming optical signal in the given wavelength range. In the input fiber there should be only one wavelength from a specific range. The switching time is in the range of milliseconds. (d) splitter. 2e) converts a wavelength from the input fiber to a required wavelength and sends it to the output fiber. All wavelengths from the input fibers are sent to the output fiber. (k) N × N AWG. Optical components: (a) multiplexer. 2j) can switch all wavelengths with its “x” or “=” state. Fig. It is controlled by the electric signal in nanoseconds.172 Fig. Fig. All wavelengths from the input fiber are sent to the output fibers. It has a single input fiber and several output fibers. which delays the optical signal. An erbium-doped fiber amplifier (EDFA. (e) fixed wavelength converter. introduces considerable noise to the optical signal. (j) 2 × 2 switch. the EDFA distorts the optical signal less. A coupler (Fig. 2i) is a piece of an optical fiber. which. An optical delay line (ODL. An optical gate can be implemented with a semiconductor optical amplifier (SOA).

there are two main ones. that we describe in some detail in the following two subsections. An optical packet switch can have an optical delay line for delaying packets for a number of time slots. One of them is called the broadcast-and-select switch. Each input fiber can carry up to as many wavelengths as there are output fibers. The input block of the switch extracts headers and interprets them. and the current state of the switching fabric. The input and output fibers carry one wavelength only. 4.173 An arrayed waveguide grating (AWG) element has a number of input fibers and a number of output fibers. Next the controller configures the switching fabric based on the header data. A general model of an optical packet switch Among many proposed types of optical packet switches. and at the right output fiber. where each wavelength is sent to a specific output fiber. Splitters broadcast the input signals.1. the other one is based on the AWG. buffer 1 input block N optical packet switch output block 1 N controller Fig. . A switch can be divided into three interconnected blocks: the wavelength encoder. 3. Each output fiber can carry a single wavelength. the space switch and the wavelength selector. The output block adds headers to payloads. 3. so that the packets leave the switch after a required number of time slots. and then the optical gates select the packets appropriately (they block them or let them through). Optical packet switches A general model of an optical packet switch is shown in Fig. 4) has N input fibers an N output fibers. The broadcast-and-select switch The broadcast-and-select switch (Fig. 4.

2. Each input fiber and each output fiber carries n wavelengths. The broadcast-and-select switch The wavelength encoder has N input fibers. and a multiplexer. The controller should not allow more than one packet to be sent through an output fiber in a time slot. The input wavelength converter has N demultiplexers and N × n tunable wavelength converters. the AWG element. Each input fiber carries a single wavelength. 4. Packets that arrive at the input fiber of the space switch are broadcast by the splitter to optical delay lines which delay the packets by a number of time slots from 1 to K. It has N input fibers and N output fibers. A block selects a packet to leave through the output fiber. The packets from the optical delay lines are broadcast again. packets can be sent to the required output fibers after the required number of time slots. The wavelength selector is built of N blocks: one block for each output fiber. The advantage of this switch type is the multicast ability: one input packet may leave the switch on multiple output fibers after various delays. depending on the wavelength of the packet. The wavelengths are multiplexed into a single fiber which is connected to the space switch. Each tunable wavelength converter receives up to one wavelength . Packets that were let through the optical gates are then sent to the output fibers through combiners.174 1 1 N 1 N 1 × ODL K × ODL K N K N 1 1 1 N K N 1 1 1 Fig. 4. which is converted by the fixed wavelength converter to a specific wavelength that does not change. A block is built of a demultiplexer. By controlling the optical gates. The disadvantage of this switch type are large power losses of optical signal and the introduction of noise. but this time to optical gates. 5). A switch based on AWG A switch based on AWG has the AWG element as the switching fabric (Fig. even on different wavelengths. and the output wavelength converter. The switch has three blocks connected in sequence: the input wavelength converter. which are the input fibers of the switch. N optical gates.

a wavelength is converted by the fixed wavelength converter to one of the n wavelengths. Unfortunately. as mentioned in section 3. two basic and useful buffer types are the circulating buffer. Optical buffers are practically and critically reviewed in [35]. The function of the AWG element is. For optical packet switching there has been even proposed a buffer based on electronic memory [34]. There are a number of types of optical buffers and ways of controlling them [33. so that the packet is sent to a specific output fiber of the AWG element. A switch based on AWG from a demultiplexer. and the cascade buffer: . 5. The optical delay lines alone are not buffers. However. The wavelength converters are tuned to a specific wavelength for each packet in each time slot. 5. 42]. 9]. The overview of optical buffers and the ways of using them with optical packet switches is discussed in [18. The output wavelength converter has N × n fixed wavelength converters and N multiplexers. there are switches needed. taking into account also slow light buffers.175 1 1 n Nn × Nn AWG 1 N n 1 1 n 1 N n Fig. After leaving the AWG element. Buffering One of the main technical problems of OPS is the lack of optical RAM memories. which will direct packets to optical delay lines. which hinders the application of the traditional store-and-forward routing. and regardless of the input fiber. and then multiplexed to an output fiber for the WDM transmission. where packets are buffered in large numbers and sent when possible [39]. ODL buffers are physically large and their capacity expressed in the number of packets is small. Optical buffering is limited mostly to optical delay lines (ODL). sending an optical signal to a specific output fiber based only on the wavelength.

an optical packet switch with its optical buffer is capable of storing only a few packets. Conclusion Optical packet switching was envisioned to be the technology of the next generation optical network. 6b) is built of several ODLs and 2 × 2 switches connected in a cascade. An electronic switch currently in operation has large electronic memory. Optical packet switching should . 38]. (b) cascade • The circulating buffer (Fig. but it still promises to be realized in the future. but a buffer with an ODL of 2 meters has been proposed too [29]. and distorting the optical signal less in comparison with the circulating buffer. • The cascade buffer (Fig. but can also reorder them. In contrast. The ODL usually has a few hundred meters. The buffer can delay more than one packet.176 Fig. which is usually undesirable. Optical buffers: (a) circulating. The switch directs a packet into and out of the ODL. Therefore the buffer can delay packets for a given number of time slots with the packet passing through a smaller number of 2 × 2 switches. and is capable of storing millions of packets. 6. 6. 6a) is built of an ODL and a 2 × 2 switch. the main problems still are the lack of RAM memories and the expensive wavelength conversion and regeneration. This vision has not been realized yet. then small optical buffers should suffice resulting only in a network performance drop of about 10% [1. A packet usually cannot circulate more than 20 times because of the optical noise introduced by the 2 × 2 switch. If the TCP (Transmission Control Protocol) traffic keeps being the predominant data traffic in the optical core networks. The subsequent ODLs can delay packets in the number of time slots equal to the powers of two. While optical packet switching offers well-established and functional switch architectures. The number of packets that can enter the ODL depends on the length of the ODL and the switching time of the 2 × 2 switch.

Chlamtac. Le Sauze. S. 3. Berde. Pujolle. 10. vol. IEEE Transactions on Parallel and Distributed Systems. C. pp. pp. IEEE Communications Magazine. Henry.B. 41.T. Mahony. G. pp. J. Bowers. L. Develder. Wang. M. L. Suh Chang-Jin: Multibuffer delay line architectures for efficient contention resolution in optical switching nodes.177 start to be deployed in operational networks when it becomes a commercially-viable alternative to the electronic hardware. vol. B. Blumenthal. B. M. SIGCOMM Comput. B. IEEE Communications Magazine. Jones. 9. Hansen. Andonovic. Kloch. A. Pickavet. IEEE Communications Magazine. pp.S. Bonaventura. Rangarajan. Fumagalli. vol. W.L. Guillemot. 7. vol.. J. Dembeck. P. Gravey. Callegati. Bregni.J. I. Chiaroni. A. A. R. A. 1245-1259. S23-S29. L. IEEE Journal on Selected Areas in Communications. 1113-1121.L. 9. RFC 2475. IEEE Transactions on. G. Ciavaglia: Distributed aggregation in all-optical wavelength routed networks. Mortensen. Rev. F. Weiss: An architecture for differentiated service. Vegetti: Architectures and performance of AWG-based optical switching nodes for IP networks. vol. IEEE Communications Magazine. Neri. Trowbridge: Optical transport network evolution: hot standardization topics in ITU-T including standards coordination aspects. 5. Krahenbuhl. Guido Appenzeller. 7. Carlson. Cerroni. Dittmann. G. S. Leligou. Raffaelli. vol. L. 7. Stavdas. F. R. 10. Feb 2003. July 1995. pp. 4. 3. vol. 4. Nick McKeown: Sizing router buffers. Cruz: Bounds on maximum delay in networks with deflection routing. 12. 7. vol. E. Chiaroni. 46. K. Sep 1998. M. Rau. T. Renaud. September 2002. A. J. Davies. pp. 7. G. 48. IEEE Journal on Selected Areas in Communications. pp. 8. 18061810. Commun. 8. F. 34. E. 60-65. C. N. References 1. IEEE Journal on Selected Areas in Communications. Callegati. Wei Wang. D. 2. 36-43. 40. Zucchelli: Transparent optical packet switching: network architecture and demonstrators in the KEOPS project. P. D. El-Bawab. 12. N. Proceedings of ICC 2004. Communications. pp. Dec 2000. N. D. Eilenberger: The European IST project DAVID: a viable approach toward optical packet switching. S. Blake. Corazza. S. Danielsen. A. Hsu-Feng Chou. Schilling. Talneau. Janz. M. 124-131. 6. G. 6. G. Le Sauze. 1026-1040. 281–292. C. pp. P. September 2003.E. vol. December 1998. 11. Bostica. 2. M. Koerber. pp. Black. October 2008. S. June 2004. vol. W. August 2007. C. 21. D. N. L. W. 724-732. Isaac Keslassy. 21. M. Bouabdallah: Sub-wavelength solutions for next-generation optical networks. vol. M. Z. Dotaro. Bouabdallah. 45. N. September 2003. 2004. 16. pp. Renaud. Brassil.N. 2089-2098. . Gambini. Rafel. Poulsen: Optical signal processing for optical packet switching networks. Pattavina. Jong-Dug Shin: Optical packet switching in core networks: between vision and reality. Sole-Pareta. I.

20. Marciniak: Droga do przyszłych całkowicie optycznych sieci pakietowych – któr˛ edy? Przeglad ˛ Telekomunikacyjny. Zouganeli: Optical label switched networks: laboratory trial and network emulator in the IST-STOLAS project. Politi. Jingyi He. 4. Hunter. 45. 28. LXXVI. vol. A. 31. vol. Shami. P. R. Nejabati. pp. 505–521. A. 2003. 4. HPSR 2003. Pickavet.T. November 2008. IEEE Communications Magazine. pp. Gumaste. Communications Magazine. Jan 2001. September 2000. 2006. D. Feletig. Gaudino. 1. J. 43-51. Jennen. Biswanath Mukherjee: Optical WDM networks. M. 12. Olmos. Mellia. C. 16. pp. 22. S. Journal of Lightwave Technology. 2003. pp. vol. Khalil. vol. 64-73. M. 58-65. 14. Chia. J. IEEE Communications Magazine.178 13. C. 25. E.-H. E. Monroy. pp. J. 170-174. pp. 166-170. I. vol. 4. 12. A. Journal of Lightwave Technology. vol.V. C. Dec. 44. IEEE Communications Magazine. IEEE. M. 2006. 36-42. Ghani. 17. December 2006. vol.K.C. Chlamtac: Light-trails: a novel conceptual framework for conducting optical communications. June 2003. 251-256. 26. pp. F. 38. D. 24. P. A. N. Andonovic: Buffering in optical packet switches. vol.M. 2004. van Berkel. 23. Koonen. Przeglad ˛ Telekomunikacyjny. I. D. 1. Qing Liu. IEEE Communications Magazine. vol. Simeonidou: Future optical networks. Lankford. 16-20. Aug. . D. P. Journal of. Gauger. vol. 12. Dec 1998. Demeester: Hybrid optical network architectures: bringing packets and circuits together. Feb 1993. E. 2006. pp. Henning: Contention reduction in core optical packet switches through electronic traffic smoothing and scheduling at the network edge. 11. Przeglad ˛ Telekomunikacyjny. pp. Lu. 4684-4696.J. R. Listanti. Neri: Exploiting OTDM technology in WDM networks.J. O’Mahony. Aug. 27. 9. D. A. Assi. Springer. 18. I. 44. Communications Magazine. 24. Gary Chan: TCP and UDP performance for Internet over optical packetswitched networks. Proceedings of INFOCOM 2002. IEEE. Jajszczyk: Transport sygnałów w sieciach nowej generacji. 54-61.V. Sabella: Architectural and technological issues for future optical Internet networks. 21. Mouftah: Photonic fast packet switching. M. C. E. A. 1822-1831. Gumaste. J. T. 3. Hunter. June 2002. V. 2. H. pp. 2081-2094. Klonidis. Lightwave Technology. vol. LXXVI. 2006. Green: Progress in optical networking. M. pp. Eramo. I. R. 46. Computer Networks. 2001. 4828-4837.T. Benhaddou: Value-added services in next-generation SONET/SDH networks. vol. 39.D. Proceedings of the Workshop on High Performance Switching and Routing. pp. 8. vol. 82-92. vol. van Breusegem. Leonardi. Kuhn. Breusegem. 8. 15. pp. pp. A. J. Jajszczyk: Ewolucja sieci stacjonarnych. 24. Jajszczyk. M.K. 16. Peucheret. M. 19. Z. pp. LXXIV.

February 1999. thinlayered. 1185-1187. B. 33. Suzuki: Photonic random access memory for 40-Gb/s 16-b burst optical packets. April 2004. vol. Yoo. vol. 9. pp. 9. Takahata. Rajiv Ramaswami. Varvarigos: Multiple-input-buffer and shared-buffer architectures for optical packet. 2005.G. A. IEEE Communications Magazine. 2009. Lisong Xu. 4. IEICE Transactions on Communications. R. S30-S36. 44. 3. pp. Yiannopoulos. T.B. vol. IEEE Communications Magazine. Nakahara. Rev. pp. IEEE Photonics Technology Letters. Burmeister. Kumar N. August 2004. vol. G. K. S. Journal of Lightwave Technology. K. Zhong: Photonic packet switching: an overview. September 2006. Mukherjee. extensible. vol. Kondo. Carla Raffaelli. Dixit: A unified study of contention-resolution schemes in optical packet-switched networks. Fei Xue. Morgan Kaufmann. 243-252. S.G. January 2001. 4655-4673. vol. Journal of Lightwave Technology. Commun. 31. 978-985. D. 40. Bjorlin. Shacham. vol.. Yokogawa Corporation. vol. 132-139. 3. S. N. December 2006. 39. SIGCOMM Comput. Bergman: A modular. 2. I. pp. 24. Mukherjee. IEEE Communications Magazine. 25. Photonic Network Communications. pp. 35. Journal of Lightwave Technology.J. B. I. May 2003. 43. J. D. pp. pp. Sivarajan. 32. Park. H. B. 3. 24. 75-78. 44. Ramaswami: Optical networking technologies: what worked and what didn’t. 672-683. E. 39. The website of the Yokogawa Corporation.J. June 2007. K. Dixit: Electrical ingress buffering and traffic aggregation for optical packet switching and their effect on TCP-level performance in optical mesh networks. Mitra: Light core and intelligent edge for a flexible. 4. Yasui. . 41. pp. Saniee. S. pp.J. pp. 136-142. Widjaja. 35. Vlachos. September 2002. W. and cost-effective optical transport network. March 2003. S.F. scalable. 25. 40. S. Paolo Zaffoni: TCP performance in optical packet-switched networks. E. Journal of Lightwave Technology. 4468-4492. 41. Shun Yao. 3rd edition. R. Small. 19-20. Tucker: The role of optics and electronics in high-capacity routers. H. IEEE Communications Magazine. 254-264. Giles. 11. S. 38. 34. 12. April 2007. 12. pp.E. 82. Nick McKeown: Part I: buffer sizes for core routers. 42. R. vol. R. 30. S. R. Yoo. Tucker.and burst-switching networks. Takenouchi. vol. 36. 1379-1389. Perros. pp. K. H.A. vol.B. Journal of Lightwave Technology. Takahashi. 21. 6. vol.B. 37. pp. 1. T. 16. and transparent optical packet buffer. Galen Sasaki: Optical networks: a practical perspective. pp.179 29. 5. Rouskas: Techniques for optical packet switching and optical burst switching. May 2006. Damon Wischik. Shun Yao. H. Yoo: Optical packet and burst switching technologies for the future photonic Internet. Bowers: 40-Gb/s optical buffer design and simulation. 66-72. vol. 4th International Conference on Numerical Simulation of Optoelectronic Devices. December 2006.

1728-1741. Yoo. 3) i bardziej szczegółowo sa ˛ omówione przełaczniki ˛ typu "rozgłaszanie i wybieranie" (rys. 7. December 2000. 21. co sprawia. Jue. Bansal. B. V. z ˛ pakietów powinno sprosta´ c jeszcze wi˛ ekszym wymaganiom. 12. z ˙ e optyczne przełaczanie iczne przełaczanie ˛ pakietów jest ulepszane. A artykule jest opisany aktualny stan wiedzy z dziedziny optycznego przełaczania ˛ pakietów z wyszczególnieniem problemów technicznych. 46. pp. 6). Na rys. ale mimo tego nie została ona jeszcze zastosowana w sieciach operatorów telekomunikacyjnych. T. Hui Zang. 5). G. Optyczne przełaczanie ˛ paki˙ etów ma wi˛ ekszy potencjał niz elektroniczne przełaczanie ˛ pakietów i dlatego nad ta ˛ ˙ y jednak pami˛ ˙ e elektrontechnologia ˛ ciagle ˛ prowadzone sa ˛ prace badawcze. K. Boyer. Goncher. 4) i przełaczniki ˛ oparte ˙ e problemy techniczne na krotnicach falowych AWG (rys. Journal of Lightwave Technology. J. Nast˛ epnie omówiona jest ogólna budowa przełaczników ˛ optycznych (rys. Nalez eta´ c. K. vol. Nady. pp. 2. Jing Cao. Y.180 45. IEEE Journal on Selected Areas in Communications. S. Mukherjee: Capacity allocation and contention resolution in a photonic slot routing all-optical WDM mesh network. Fei Xue.B. J. Omówione zostały takz zwiazane ˛ z optycznym buforowaniem pakietów (rys.J. Przeglad ˛ optycznego przełaczania ˛ pakietów Streszczenie Technologia optycznego przełaczania ˛ pakietów jest tematem prac badawczych od około dwóch dekad. Okamoto. vol. 1041-1051. S. aby mogło by´ c zastosowane w sieciach operatorów telekomunikacyjnych. Taylor. Zhong Pan. 18. 1 pokazany jest schemat sieci optycznej. September 2003. Akella: High-performance optical-label switching packet routers and smart edge routers for the next-generation Internet.P. Minyong Jeon. której w˛ ezły sa ˛ zbudowane z elementów optycznych symbolicznie przedstawionych na rys. . We wprowadzeniu omawiane sa ˛ podstawowe sposoby komunikacji w sieciach optycznych. Kamei.