Software Engineering 4C04 Research Project:
The Joys and Sorrows of Optical Routing
Friday, March 23,2001
In the last few of years much research and many resources have been devoted to the development of optical network technologies. The main goal behind this focus is the actualization of the purely optical network. An essential but as of yet, unattained, component of this goal is the development of optical routing technology.
This paper is intended to give the reader an insight into and an understanding of this emerging hardware technology. The layout of this paper will bring the reader through the necessary background information such as what a router is and how it is currently implemented through to information needed to understand optical routing. As well, information on the current state of development and what to expect in the future will be given.
What is routing?
The Internet is a worldwide collection of computer networks, which relies on the TCP/IP protocols to transmit digital information between networks and individual computers. These protocols specify, among other things, what digital information should go where.
The actual physical devices that direct this digital traffic to its intended destination are routers. While switches or other interconnects are used within individual networks, routers are used to direct traffic between networks. And so, each router is connected to two or more networks. Inside each router, software utilizes the TCP/IP protocols to send, or route, packets of digital data to the appropriate destination.
Routing can be a very complicated task and ideally, the packets would be directed by selecting the optimal path of transmission though the monitoring of several factors such as the distention address, network load, packet length and the type of service provided by the packet. In many cases, routers simply rely on information provided in the packet headers (a section of the packet which specifies information about the data contained in the packet) to select routes based on fixed shorted path assumptions.
Electrical vs. Optical Routing
The connection to a specific router can be either optical cables or one of several types of electrical, copper based cables. Electrical cables use voltages to encode and transmit information while optical cables are made of glass fibers and use light to convey information.
The router takes the header from the packet and chooses the path based on information found within. This operation is relatively straightforward when the type of connection is an electric signal based cable were a specified number of bits, the header, can be separated from the rest of the packet and read by the routers processor. This task becomes more difficult when the type of connection to the router is an optical cable. In this situation, before the information contained in the header can be processed, the packet must be converted from optical to electrical signals. This is due to the fact that today's processors work in the electrical domain and cannot directly process optical signals. Before this information can be sent on, the signal must be converted back into the optical domain.
The eventual goal of research into all-optical routing is to develop a router that would not have to do this optical-electrical-optical conversion. If this was accomplished, the purely optical network in which end-to-end service is provided without ever converting the optical signal into the electrical domain could be implemented. There are several proposed methods to achieve this and many resources are being allocated towards its development. Advances are being made but no one has yet achieved this all-optical router. Problems with this all optical implementation are discussed in the later section titled "Barriers to all-optical routing".
There are many benefits gained from the use of optical fibers in the place of electrical cables. Optical cables are less affected by the distortion of signals due to electro-magnetic interference caused by close proximity to electric devices. This interference can introduce errors in the information carried by electrical cables. The amount of power required to generate these optical signals is less, leading to a reduction in transmission costs. The degeneration of optical signals is much less then electrical signals due to the amount of resistance in the medium. This means that an optical signal can travel for far greater distances without the need for regeneration. As well, the amount of information that can be carried through an optical cable is much greater. All of these factors lead to a much-reduced cost for the carriers, making optical cables a very attractive, and widely used, transmission medium.
Since many networks already use optical cables as the transmission medium, the advantages of an all-optical router are clear. As any one packet may have to path through multiple different physical networks on its path from sender to receiver, it is likely that it will have to be routed many times. Each time that the packet is routed, it will have to be converted to an electrical signal to be processed and then back to optical signal to be sent on. This conversion process takes time and as the distance and the number of times the packet must be routed increases, the time the packet takes to arrive at its destination will also increase.
In today's market, which thrives on high bandwidth and transmission speed, a faster network is highly sought after. The latency introduced into the system by optical-electrical-optical conversions at network routers can have the effect of lowering the bandwidth of the connection. Thus the introduction of all-optical routing is advantageous in that it will, among other benefits, increase transmission speeds along optical cables.
Barriers to all-optical routing
There are several problems that must be overcome before the goal of all-optical routing can be achieved. If the entire packet is to remain in the optical domain, how does the router read the destination information and ensure that the packet is sent along the correct path? There are two proposed methods for intelligently directing packets throughout optical networks. One solution would require the development of an optical processor that would be able to process the incoming optical signal in much the same manner as the current electrical processors process electrical signals.
Another solution is somewhat a hybrid solution but can be considered all-optical by the fact that the actual packet remains in the optical domain while only the header will be replicated as an electrical signal. This solution requires the extraction of the header from the body of the packet and its subsequent conversion into the electrical domain where an electrical processor can effectively route the optical packet. Since the bit size of the header is relatively small the electrical process would be able to keep up with the optics.
Another problem arises when we look at how multiple incoming packets are handled. With an electronic signal it is easy just to store this information into a buffer and read it out when the processor is free. Now if this signal is not going to be converted into the electrical domain, how does the router handle simultaneously incoming packets? The proposed solution for this, which has been developed in the laboratory, is to loop the lightwave around a closed connection until the processor is ready for it. This solution is aided by the speed of which all optical routing can be done limiting the size needed for this optical buffer.
What does the future hold?
Currently, the implementation of all optical switches and add/drop mutliplexors are starting to enter the market and be used to direct intra-network traffic. These switches use several different methods from mechanical motors aligning fiber cable ends to tiny silicon mirrors reflecting light waves, or the use of liquid crystal matrices to open and close connections between opposing cable ends. This however is a much simpler task to implement then the actual intelligent selection of routes through many different possibilities.
As for the all-optical routing problem, there currently exists many research papers and proposed solutions but currently, as well as for the immediate future, the purely optical network is likely to remain confined to the laboratory. There is much debate from experts as to what will be accomplished and when it will be accomplished. So the actual timeline before we will see readily available all-optical routers on the commercial market or if we ever will, is unclear. What is certain is the ever-increasing internet traffic and demand for bandwidth. As this trend continues the need for ultra-fast packet routing between optical networks will become more pronounced. Some say this need cannot be adequately fulfilled by conventional electrical signal based routers, others say that electronic will never be entirely eliminated. Whatever the case, speculation has not stopped billions of dollars from being invested to drive the search for an all-optical routing solution.
No matter how much capital is being spent on the development of all optical networks, there are many practical barriers to their attainment. The purpose of using these all-optical networks would be to improve the speed and performance of a network while reducing the cost. Therefore it must be proven that there will sufficient benefits in these areas to drive the carriers to invest in this new equipment, when/if it is developed. Therefore I see this as a viable area to invest resources since the potential benefits are enormous, but actual widespread market deployment of these purely optical technologies is, most likely, a long way off.
1. Comer, D.E. Internetworking with TCP/IP. Prentice Hall, New Jersey. 2000.
7. Paul Toliver, Ivan Glesk, Robert J. Runser, Kung-Li Deng, Ben Y. Yu, Paul R.
Prucnal; Routing of 100 Gb/s Words in a Packet-Switched Optical Networking
Demonstration (POND) Node.Journal of Lightwave Technology,Vol 16, No 12,
8. R.J. Runser, K.-L. Deng, P. Toliver, I. Glesk and P.R. Prucnal; A highly scalable
OTDM router using a computer controlled time slot tuner with picosecond resolution;
Department of Electrical Engineering, B-204 E-Quad, Olden Street, Princeton
University, Princeton, New Jersey 08544, USA