Read OPSforMAN_R4 text version

All-Optical Packet Switching for Metropolitan Area Networks: Opportunities and Challenges

Shun Yao1,4, S. J. Ben Yoo1, Biswanath Mukherjee2, Sudhir Dixit3 1 Department of Electrical Engineering, University of California, Davis 2 Department of Computer Science, University of California, Davis 3 Nokia Research Center, Boston 4 Contact email: [email protected]

Keywords: IP-over-WDM; metropolitan area networks, optical networking, packet switching, photonic switching systems 1. Introduction The past few years have witnessed fundamental and dramatic evolutions in network traffic demand and networking technologies. The network operators, carriers and service providers are quick to embrace new technologies to meet the ever-growing user demand for more bandwidth, faster provisioning and richer sets of functionalities. As a result, the network architectures are evolving constantly with new technologies such as Dense Wavelength Division Multiplexing (DWDM) and Multi-Protocol Label Switching (MPLS). The Open System Interconnection (OSI) network layers, defined decades ago, are no longer applicable in the new networking environment. Typically, the intelligence inside a network, for example the routing information distribution and data forwarding, is implemented electronically, while the actual transport of bits is achieved optically. This is because electronics is well developed to carry out sophisticated processing, monitoring and switching, while optics is well developed for transmitting data. As the electronic circuits achieving moderate advancement in speed and optical technologies demonstrating remarkable advances, it appears to be an inevitable trend that more and more intelligent functionalities, such as switching, will be implemented in the optical layer. The network evolutions have always been induced by disruptive technologies, as we have seen Synchronous Optical Network (SONET), Asynchronous Transfer Mode (ATM) and MPLS. It is our belief that all-optical packet switching will eventually emerge as one of such technologies, which will significantly impact on next-generation networks. A number of studies have discussed various implementation and architecture issues. This paper not only attempts to briefly discuss these issues, but also to provide a higher-level view of how all-optical packet switching can be used to build a m flexible ore next-generation metropolitan area network (MAN). 2. All-optical Packet switching for MAN 2.1 Network requirements In an end-to-end connectivity picture, the current networks are consisted of three major segments: the access network, MAN, and the backbone Wide Area Network

(WAN). The access network is responsible for collecting end-user traffic and is usually of less than tens of miles in its extent. Examples of access networks include intra-building Ethernet and networks operated by local Internet service providers (ISP's). The MAN is responsible for transporting traffic between different access networks and route part of the traffic onto the backbone WAN network. The MAN usually does not exceed a few hundred of miles in size. The backbone WAN interconnects MAN's that are typically hundreds to thousands of miles away. Optics plays a key role in transmission in the backbone WAN. These three segments of networks complete the end-to-end delivery path for user data, and each of them has different characteristics. The access network does not need to offer high bandwidth as MAN or WAN, since it only deals primarily with low-end users. The main consideration for building an access network is to keep the cost low. Therefore most access networks are based on electrical media (copper wire) or passive optics (low-cost fiber systems without regeneration) with a simple media access control (MAC) protocol. In MAN, the clients are different access networks and high-end users (such as financial institutions and large ISP's, who require reliable, high bandwidth connectivity nationwide). The MAN needs to provide a large variety of service qualities, such as best-effort based connection-less datagram delivery and/or Quality of Service (QoS) based connection-oriented virtual circuits. It should also offer scalability to accommodate the rapidly growing number of access networks. Another issue MAN has to cope with is the time-dependency of traffic patterns. The heavy load of the network may move from the business district in downtown during working hours to the residence area in the suburbs in the evening. Part of the MAN traffic will travel inside the same MAN, the rest will have to be transported outside the MAN to another MAN through the backbone WAN. The WAN may have varying extent of link lengths, typically long (>100 miles) but some times shorter in densely populated areas where a number of MAN's may reside closely with each other. The traffic on the backbone WAN, compared with that of MAN, is aggregated, groomed, relatively more static and predictable. High bandwidth light paths are used to build connectivity for WAN's. Numerous recent start-up companies target MAN as their primary market because of its lack of a fixed network architecture and its accelerated growth. Future MAN's must provide users with high bandwidth, together with different granularities of bandwidth. Meanwhile, the emergence of various Internet applications and ever-growing number of users demand a future-proof MAN with excellent scalability, multi-protocol support and rapid provisioning capabilities. Typical MAN's of today are based on ring topology and SONET technology (Fig. 1). Connectivity is established by using SONET add/drop multiplexers (ADM's), and digital crossconnects (DXC's) in the inter-connected ring case. SONET equipment provides Time Division Multiplexing (TDM) subchannel switching within a wavelength. SONET terminals, ADM's and DXC's all require opticalelectrical-optical (OEO) conversion and time multiplexing/demultiplexing. Lately, a few companies started manufacturing optical crossconnect-based products, which are able to perform wavelength switching without time demultiplexing the data carried by the wavelength. The recent Multi-Protocol Lambda Switching (MPS) and Generalized Multi-Protocol Label Switching (GMPLS) activities in IETF further suggest using MPLS protocols to expedite wavelength (and therefore end-to-end connectivity when combined with IP routers) provisioning. Although optical crossconnects (OXC's) and optical add-

drop multiplexers (OADM's) can circumvent processing the bits and maneuvering them in the time domain, the granularity cannot be finer than a single wavelength. Therefore, the flexibility of such MAN's constructed with OXC's and OADM's is limited. 2.2 Promises of All-optical Packet switching All-optical packet switching, while combining high throughput and packet level switching, appears to be a good candidate for MAN application. Although most of the past investigation was carried out in research institutions and we are still far from any definitive architectural design of an all-optical packet switching node, there are two common features that an all-optical packet switch should have: a) The switching takes place in the optical domain without OEO conversion. Alloptical switching ensures higher throughput and less power consumption. b) The switch should be able to perform packet level switching. This is what differentiates an all-optical packet switch from an OXC. Theoretically, the switch fabric should also be able to switch with more coarse granularities such as wavelengths. Although the architectures proposed by different research bodies vary to a certain extent, these two main features should remain the same. Fig. 2 shows a generalized illustration of an all-optical packet switching node. The input stage is responsible for pre-amplification when necessary, reading the packet headers and packet alignment if so desired. Delay lines are used between the input stage and the switch unit to give enough time to the header-processing unit to configure the switch before the packets enter the switch unit. The header-processing unit reads the header and accordingly configures the switch. The switch unit, besides switching the packet to the desired port, also carries out contention resolutions. The output stage might perform header re-writing and power amplification if necessary. Before we further discuss the issues involved in the all-optical packet switch architectures, it is important to first investigate what advantages could be obtained by incorporating all-optical packet switching into the future networks. The ubiquity of IP routers today is an example that packet switching is more capable to adapt to the fast changing networks, because of its high flexibility. An IP network can carry connection-oriented (TCP) traffic as well as connection-less (UDP) traffic, and be built upon various lower layer technologies (ATM or SONET). Similarly, all-optical packet switched networks will be able to carry traffic from different upper layers. The limitations of IP include: a) the simple destination-based routing imposes limitations on the routing functionality and traffic engineering and b) switching in the electronic domain is limited by the speed and the power consumption in the electronics. MPLS successfully solves the first issue by using exact-match forwarding algorithm and constraint-based routing. All-optical packet switching will further remove the OEO conversion bottleneck and offer a combination of high throughput, considerable flexibility and a rich set of routing functionality similar to that of MPLS.

The future MAN will be a very dynamic network, consisted of a large variety of users. Different users will have different service requirements. For example, the corporate clients are likely to have traffic mainly generated by IP, and therefore require high bandwidth, IP-based connectivity between major offices. High-end clients, such as financial institutions, will likely to request highly reliable, connection-oriented connectivity. Campus networks and local ISP's are likely to generate IP traffic mainly consisted of web traffic, therefore request best-effort IP-based connectivity. All-optical packet switching will be able to provide a large variety of services to meet these different needs, because of its capability to switch with different granularities. An all-optical packet switch, since it does not read the actual bits in the payload data, can stay in a given switching state for an arbitrary amount of time. This unique feature enables us to provide a large range of services. For example, for campus networks or low-end ISP's, we can provide best-effort service; for corporate clients we can provide reliable virtual circuit (similar to ATM) service; and for high-end clients we can provide a whole protected light path. Another advantage of using all-optical packet switching for MAN is that we can use the existing protocol suites for routing and signaling. The on-going work at IETF is extending the MPLS protocol suite to build a seamlessly integrated network of IP routers running with OXC's [1]. It is possible that in the future the generalized MPLS will be running on both LSR's and OXC's. Since an all-optical packet-switch can function like both LSR and OXC, it will be easy to run the same protocols and integrate them with the existing equipment. 2.3 Internetwork interfaces The aforementioned characteristics of all-optical packet switching make it an excellent candidate for MAN. But to complete the end-to-end connectivity, we also need to consider how the all-optical packet switched MAN interfaces with the rest of the networks. On the client side are the access networks and high-end users, whose networks may be built with IP routers or ATM switches. On the other side, the MAN will have one or more egress nodes, which are connected to the high-speed backbone WAN. The client side of an all-optical packet switching node is depicted in Fig. 3. The client interface should be able to process layer 3 or layer 2 traffic or, connection requests when a wavelength is requested, based on different protocols such as IP, ATM and SONET. In the case of IP packets, the client interface will first look at the IP header, assign an optical header to it according to the routing table, and transmit the packet to the next all-optical packet switching node. To be able to fully exploit the capacity of all-optical packet switching, the client interface should be a high performance edge label switch router-like device with optical interfaces in order to aggregate user traffic and switch it at optical line speed. There are two options as to how to send packets from the all-optical packet switched MAN to the backbone WAN. The first option is to have the egress node perform OEO conversion, read the layer 3 or layer 2 headers, and then transmit the packets onto the established light paths through the backbone (Fig. 4a). This is similar to the overlay

model in MPS, where the inside of the optical crossconnected network is opaque to the outside. The other option, as shown in Fig. 4b, is to utilize label stacking to deliver the packets to the desired all-optical packet switched node or LSR on the other side of the backbone. This will require the label binding information from several network segments to be propagated through the whole delivery path during the label switched path (LSP) setup period. In the example in Fig. 4b, a packet has to go from node X.1 to node Z.3. Network X and Z are all-optical packet switched MAN, while network Y is the optical crossconnected backbone. Network Y is assumed to be running MPS and label binding information is implicit in the wavelength index b. During the LSP setup phase, both network Z and network Y will send the label binding information to node X.1. (In the context of all-optical packet switching we assume network segment-wide unique label because label swapping in optical domain is a nontrivial task, unlike in the electronic domain.) Node X.1 will stack the labels from all three networks (label a from network Z, label/wavelength index b from network Y and label c from network X), form an optical header and assign it to the packet. When the packet is leaving network X, the egress node will pop one label from the stack and the ingress node at network Y will use the next level label (b) to assign the wavelength. At the egress node of network Y the same label pop process will take place and the packet will eventually reach the desired egress node Z.3. If optical label swapping is applicable, it will not be necessary for the label binding information to be passed to the source node. The labels will be of local significance only. 3. Architectures for all-optical packet-switching 3.1 Slotted and unslotted networks In general, all-optical packet-switched networks can be divided into two categories: slotted and unslotted. In the slotted networks, the packets have fixed size and are placed in time slots. The size of slots is larger than the packet size to allow guard time before and after each packet. Since a particular node has input fibers from different upstream nodes, the slot boundaries on these fibers are not synchronized. There will be time variation of these boundaries due to different propagation distances and temperature changes. Certain synchronization mechanisms need to be implemented to align the slots before the packets enter the space switch. Such mechanism can be realized by using switched fiber delay lines of different lengths to create desired delays with limited resolution. Slotted networks have less contentions than unslotted networks because the packets are of the same size and are maneuvered in a joint manner. This is similar to the slotted ALOHA protocol in the carrier sense multiple access / collision detection (CSMA/CD) networks. Another motivation for slotted networks is that optical buffering can only be implemented with fiber delay lines, which are strict first-in-first-out (FIFO) queues with fixed delays. In the unslotted networks, packets can have variable sizes and are not aligned before they enter the switch. The switch has to switch every packet `on the fly'. This type of networks have more contentions and, therefore, higher packet loss probability. The node architecture for unslotted networks can be simpler because there is no need for the synchronization stages. Furthermore, packet fragmentation and re-assembly are not

necessary at the ingress and egress nodes, making such networks more suitable to switch native IP packets. 3.2 Contention resolutions In an all-optical packet switched network, contention occurs at a node whenever two or more packets are trying to leave the switch from the same output port, on the same wavelength. How the contention is resolved has a significant effect on the network performance in terms of packet-loss ratio, average packet delay, average hop distance, and network utilization. In electrical packet-switched networks, contention is resolved with the store-and-forward technique, which requires the packets in contention to be stored in a memory bank and sent out at a later time when the desired output port is free. This is possible because of the availability of electronic random-access memory (RAM). There is no equivalent optical RAM available; therefore, different approaches have to be used. Meanwhile, WDM networks provide one new additional dimension, namely wavelength, for contention-resolution. Therefore, we are able to explore three dimensions of contention-resolution schemes: wavelength, time, and space. Wavelength conversion is the most effective contention resolution without incurring additional latency while maintaining the shortest path or minimum hop distance. Wavelength converters are used to change the wavelength of the packets in contention, as a result multiple packets can be sent simultaneously to the same output port. In time buffering, fiber delay lines are typically used, together with the switches, to delay the packets in contention and send them out at a later time. Since fiber delay lines are strict FIFO queues with fixed delays, they are far less efficient than electronic RAM. To be able to resolve contention effectively, a large number of fiber delay lines are usually required. Handling many fibers in a node can be a difficult task. In space deflection, packets in contention are deflected to output ports other than the desired one. As a matter of fact, part of the network is used as `buffers' to store these packets. Although packets are sent to a different node, this node should be able to route the packet toward its final destination. The disadvantage of deflection is that it introduces extra link propagation delays and causes packets arriving out of order. One alternative approach to resolve contentions, other than the aforementioned schemes, is optical burst switching (OBS)[2]. The switch granularity in OBS is burst, which is a collection of multiple IP packets with several megabytes of data. In OBS, the source node first sends out a control packet to all the switches along the route the burst will be traveling on later. After a certain offset time, the burst is sent out without getting any acknowledgement from the switches. By the time that the burst arrives at the switches, the switches will have the necessary port reserved for the burst. After t e burst h is transmitted through the switch, the port is released. By manipulating the amount of offset time between the control packet and the data burst, different class of service can be implemented. In OBS, small-sized IP packets (such as 40 byte IP packets corresponding to Internet control message protocol (ICMP) messages or 44 byte TCP acknowledgment packets) might be disfavored, since it will take numerous such packets to assemble the

burst. The difference between OBS and all-optical packet switching is mainly in the switch control software, therefore an all-optical packet switching node can also be configured to carry out OBS. 3.3 Node architectures There have been several proposed node architectures to carry out all-optical packet switching. One of the representative examples is the European KEOPS broadcast and select switch [3]. Although wavelength converters are used in this architecture, they are used, together with semiconductor amplifier gates, only to carry out the space switching and delay selection functions, instead of to resolve contentions. Contentions are resolved by fiber delay lines. Because of its broadcast-and-select nature, a large number of semiconductor amplifier gates (N×(N+K), where N is the number of switch input ports and K is the number of fiber delay lines) are used. The WASPNET architecture developed at the University of Strathclyde, UK uses tunable wavelength converters, AWG's, and fiber delay lines to resolve contentions [4]. The switch is consisted of two stages. The first stage is responsible for contention resolution, while the second stage is responsible for space switching. This architecture requires a 2N×2N AWG's, 4N wavelength converters, an N×N space switch (where N is the number of switch input ports) and a number of f iber delay lines. Both the KEOPS and WASPNET architecture assume fixed packet size. A hybrid contention resolution combining wavelength, time and space appears to outperform any single schemes [5]. Further more, the use of wavelength conversion and deflection routing will help reduce the number of fiber delay lines. Fig. 5(a) shows the node architecture currently under for contention resolution. The node has a rather simple configuration, with wavelength converters placed on the output ports of the space s witch and a number of fiber delay lines. Our simulation results in [6] indicate that with four fiber delay lines, overall packet loss rate can be kept below 1% with transmitter load less than 30%. The network topology used in the simulation is shown in Fig. 6(b). We assumed unslotted operation with exponential packet size distribution. For any all-optical packet-switch to be actually practical, the packet loss rate in a steady network state should be well below 1%, which is the upper limit for any realistic TCP connections [7]. Moreover, for the all-optical packet-switch to be competitive with the upcoming terabit high-speed IP routers or Label Switching Routers, it should operate beyond 10Gb/s per wavelength to actually take advantage of the optical transparency of the payload data. At 40Gb/s and up, regeneration will be necessary after the signals travels through various components in a number of switching nodes. All-optical regeneration solutions will have to be to be pursued if OEO conversion is to be a voided throughout the network. 4. Enabling technologies The core of an all-optical packet-switching node is the switch fabric. To be able to switch on packet level at multi Gb/s speed, the switch must have nanosecond switching time. Micro electronic machine system (MEMS) technology is not applicable here

because the MEMS switches can only perform millisecond switching. Potential candidates include LiNbO3 based switch elements and semiconductor optical amplifier (SOA) gate switch elements. An alternative is to use ultrafast tunable lasers; wavelength converters and static wavelength routing devices, such as AWG's, to construct a fast, large port count switch fabric. This approach can avoid concatenating multiple stages of 2x2 switch elements because of the large port-count of AWG devices. Wavelength conversion remains one key technology in contention resolutions. Although the current stage of wavelength converters is not as mature as other devices, it is only a matter of time to have off-the-shelf wavelength converters. For example, Alcatel already has a packaged all-optical, active Mach-Zehnder structure-based wavelength converter operating at 10Ghz. Parametric wavelength conversion may be of special interest in all-optical packet switching because of its capability of converting multiple wavelengths simultaneously [8]. The current research in optical RAM is still in an early stage, while there are some promising discoveries in laboratory level (e.g. the chiropticene molecular switch [9] and molecular transistor [10]). Before optical RAM becomes available, we still need to rely on fiber delay lines as optical buffers. On the software side, the all-optical packet-switched network can utilize the existing MPLS / MPS protocols with necessary extension for signaling, and IP routing protocols for routing information distribution. There is no need to re-invent another protocol layer exclusively dedicated to all-optical packet switching. 5. Summary As we continue to observe the exponential growth in data, especially IP, traffic, the metropolitan area network will have to manage a great variety of clients and meet different client requirements in terms of bandwidth provisioning and service qualities. All-optical packet switching appears to be a suitable candidate to build a flexible, high throughput and scalable MAN. Powered by the generalized MPLS protocols, such alloptical packet-switched networks can be able to provide different switching granularities, such as packets; bursts; and circuits, and support different u data formats. To be able ser to compete with high-speed IP routers and label switch routers (LSR's), the all-optical packet switch should operate beyond 10Gb/s. Optical regeneration will be necessary as the bit rate approaches 40Gb/s. A simple node architecture employing wavelength conversion, optical buffering and deflection routing is presented and simulated on an irregular mesh network topology. With the current switching and buffering technology we still have to make a trade-off between packet loss rate, network utilization and node complexity. Fast, large port-count switch fabrics and wavelength converters remain as the key enabling technologies before there is a breakthrough in optical RAM.

REFERENCES [1] P. Ashwood-Smith et al. "Generalized MPLS - Signaling Functional Description", Internet Draft, draft-ietf-mpls-generalized-signaling-01.txt, November 2000. [2] C. Qiao and M. Yoo, "Optical Burst Switching (OBS): A New Paradigm for an Optical Internet," J. High Speed Networks, Special Issue on Optical Networking, vol. 8, no. 1, Jan. 1999, pp. 69-84. [3] C. Guillemot et al., "Transparent Optical Packet Switching: the European ACTS KEOPS Project Approach," IEEE/OSA J. Lightwave Tech., vol. 16, no. 12, Dec. 1998, pp. 2117-34. [4] D. K. Hunter et al., "WASPNET: a Wavelength Switched Packet Network," IEEE Commun. Mag., Mar. 1999, pp. 120-29. [5] S. Yao, B. Mukherjee, S. J. B. Yoo and S. Dixit, "All-optical Packet-switched Networks: A Study of Contention-resolution Schemes in an Irregular Mesh Network with Variable-sized Packets," Proceedings, OptiComm 2000: Optical Networking and Communications, Richardson, TX, Nov. 2000. [6] S. Yao, S. J. B. Yoo and B. Mukherjee, "A Comparison Study Between Slotted and Unslotted All-optical Packet-switched Network with Priority-based Routing," OFC '01, Anaheim, CA, Mar. 2001. [7] T. V. Lakshman and U. Madhow, "The performance of TCP/IP for networks with high bandwidth-delay products and random loss," IEEE/ACM Trans. Networking., vol. 5, Jun. 1997, pp. 336-350. [8] S. J. B. Yoo, C. Caneau, R. Bhat, M. A. Koza, A. Rajhel, and N. Antoniades, "Wavelength conversion by difference-frequency generation in AlGaAs waveguides with periodic domain inversion achieved by wafer bonding," Appl. Phys. Lett, 68, pp. 2609-2611, 1996. [9], Also see U.S. Patent, R. R. Schumaker, number US05237067, "Optoelectronic Tautomeric Compositions". [10]

Shun Yao (S'99 [email protected]) received his B.E. in electronic engineering in 1997 from Tsinghua University, China, and is currently a PhD candidate at the University of California, Davis. He is working closely with Nokia Research Center (Boston) on optical packet switched networks. His research interests

include WDM networks, all-optical packet switching, MPLS, and various network control and management issues. S. J. Ben Yoo (S'82­M'84­SM'97 [email protected]) received the B.S. degree with distinction in electrical engineering, the M.S. degree in electrical engineering, and the Ph.D. degree in electrical engineering with minor in physics, all from Stanford University, Stanford, CA. His Ph.D. dissertation at Stanford University was on linear and nonlinear optical spectroscopy of quantum-well intersubband transitions. He is an Associate Professor of Electrical and Computer Engineering at University of California (UC), Davis, where his current research involves advanced switching techniques and cross-connects for the Next Generation Internet. Prior to joining UC in 1999, he was a Senior Scientist at Bellcore, Red Bank, NJ, leading technical efforts in optical networking research. His research activities at Bellcore included optical-label switching for the Next-Generation Internet, power transients in reconfigurable optical networks, wavelength interchanging cross-connects, wavelength converters, vertical cavity lasers, and high-speed modulators. He also participated in the ATD/MONET systems integration, the OC-192 SONET Ring studies, and a number of standardization activities. Prior to joining Bellcore in 1991, he conducted research on nonlinear optical processes in quantum wells, four-wave mixing study of relaxation mechanisms in dye molecules, and ultrafast diffusion driven photodetectors. During this period, he also conducted research on life-time measurements of intersubband transitions and on nonlinear optical storage mechanisms at Bell Laboratories and at IBM Research Laboratories, respectively. Dr. Yoo is a member of the Optical Society of America (OSA) and Tau Beta Pi. Biswanath Mukherjee ([email protected]) received a B.Tech. (Hons) degree from Indian Institute of Technology, Kharagpur, in 1980, and a Ph.D. degree from the University of Washington, Seattle, in June 1987. At Washington he held a GTE Teaching Fellowship and a General Electric Foundation Fellowship. In July 1987 he joined the University of California-Davis, where he has been professor of computer science since July 1995, and chairman of computer science since September 1997. He is co-winner of paper awards presented at the 1991 and 1994 National Computer Security Conferences. He serves on the editorial boards of IEEE/ACM Transactions on Networking, IEEE Network, ACM/Baltzer Wireless Information Networks (WINET), Journal of High-Speed Networks , Photonic Network Communications, and Optical Network Magazine. He also serves as Editor-at-Large for optical networking and communications for the IEEE Communications Society. He served as Technical Program Chair of IEEE INFOCOM '96. He is author of the textbook Optical Communication Networks (McGraw-Hill, 1997), a book which received the Association of American Publishers, Inc.'s 1997 Honorable Mention in Computer Science. His research interests include lightwave networks, network security, and wireless networks. SUDHIR DIXIT (SM [email protected]) received a B.E. degree from Maulana Azad College of Technology (MACT), Bhopal, India, an M.E. degree from Birla

Institute of Technology and Science (BITS), Pilani, India, and a Ph.D. degree from the University of Strathclyde, Glasgow, Scotland, all in electrical engineering. He also received an M.B.A. degree from Florida Institute of Technology, Melbourne. He currently heads research in broadband networks at Nokia Research Center, Boston, Massachusetts, specializing in ATM, Internet, all-optical networks, and third-generation mobile networks. From 1991 to 1996 he was a broadband network architect at NYNEX Science and Technology (now Bell Atlantic). Prior to that he held various engineering and management positions at other major companies, such as GTE, Motorola, Wang, Harris, and STL (now Nortel Europe). He has published extensively, and has 13 patents either granted or pending. He has been either conference chair, session chair, and/or on the program committees of several conferences. As an ATM Forum Ambassador, he has presented tutorials on ATM internationally. He was a guest editor for a special issue of IEEE Network on digital video dial-tone networks, published in October/November 1995, and a guest editor for a feature topic on service and network interworking in a WAN environment published in IEEE Communications Magazine in June 1996. He is a Guest Editor of a special issue of IEEE Communications Magazine, "WDM Networking: A Reality Check," to be published in March 2000. He is listed in several national and international Who's Who publications. He is an editor and light-wave series editor of IEEE Communications Magazine.

Node 6

SONET frame

Local add/drop ports

Node 1



Node 2

SONET frame header

Local Add

SONET ADM SONET ADM SONET virtual container

Node 3

Local Drop

Node 4

Node 5

Fig. 1 The SONET ring based network.

Header Processing Unit

Switch Unit

Fig.2 Generic block diagram for an all-optical packet switch.

Client interface


Aggregate access network traffic and assign optical header.

Fig. 3 Client interface between all-optical packet switched MAN and access networks.

Output Stage

Input Stage




Take traffic from backbone egress, look at layer 2/3 (ATM, IP, etc) address, and assign optical header for MAN. (a)

Network X X.1

Fig. 4 Two options to inter-connect the all-optical packet switched MAN with optical crossconnected backbone. (a) Using electrical equipment at the interface. (b) Using label stacking and end-to-end signaling.

a b c a b

Network Y



Network Z


8 9 7 2 10 11 3 5 14 4 13 12 1 6 15





Fig. 5 (a) A simple all-optical packet switch architecture employing wavelength conversion, optical buffering and deflection routing, (b) the network topology simulated.



14 pages

Find more like this

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate


You might also be interested in

Ciena's Intelligent Optical Network Deployment and Directions
New Jersey Undergraduate Academic Catalog | Information | DeVry University
17075 v#2 10.8.08