A multichannel Ethernet protocol for a WDM local area star network

Transcription

1 A multichannel Ethernet protocol for a WDM local area star network D. Rodellar, C. Bungarzeanu, H. Garcia, C. Brisson, A. Küng, Ph. Robert Telecommunications Laboratory (1), Metrology Laboratory Swiss Federal Institute of Technology - EPFL 9th IEEE WORKSHOP ON LOCAL AND METROPOLITAN AREA NETWORKS: LANMAN 98 The Banff Centre for Conferences Banff, Alberta, Canada May 1998 Session 7: Optical Networks & WDM Oral presentation: Tuesday, May 19, 2.00pm - 3:30pm (1) For further information - D.R: Daniel.Rodellar@epfl.ch; Telephone: Fax: ; TCOM - Department of Electrical Engineering CH-1015 LAUSANNE;

2 ABSTRACT Forthcoming applications will need high guaranteed bandwidth and low bounded delay. This paper illustrates the advantage of distributing total capacity over several wavelengths with an appropriate protocol. Using the benefits of the multichannel capabilities, a new Ethernet protocol based on a Wavelength Division Multiplexing (WDM) scheme is proposed. The major goals are to geographically extend the network up to several km, to obtain an adequate delay and throughput performances in order to support real Local Area Network traffic, to simplify the transmission and reception schemes, to be compatible with existing standards and finally to satisfy the high demand of the large number of users connected. A logical bus Local Area Network is implemented over a passive star physical topology. Each node has a tunable transmitter in order to choose the channel where to transmit its packets. It is also equipped with a coherent receiver that captures packets from all the different wavelengths. This reception is achieved using coherent technology. The incoming optical signal is mixed with that of an optical local oscillator, which is shared by all the hosts, and then a group of electrical filters selects each different data channel at different intermediate frequencies. WDM optical networks emerge as a viable alternative to provide Gbit/s communications. A new protocol based on Ethernet, which senses the carrier (CSMA) with collisions detection and thus requires packet retransmission, is proposed. Channel jumping is introduced on the protocol as a new dimension. Hosts choose a wavelength to transmit data and if the chosen channel is busy packets have the possibility to escape to another channel. The new parameters are the number of consecutive collisions that one host has to wait before it changes to another wavelength, and the laser tuning time, which is the time required for the laser to jump from one wavelength to another. This new protocol is compatible with Ethernet as long as those new parameters are taken from the physical layer. Network protocol behavior is first simulated by mixing long and short fixed length packets, which is more judicious than a fixed unique packet length model. A finite number of hosts generates packets with poissonian arrivals and with a distribution of packets lengths. Then the same simulations are done with long-range dependent traffic (). Simulation of 10 channels of 10 Mbit/s each and a single channel of 100 Mbit/s have allowed us to compare the transmission average delay, the packet loss rate and the throughput for the two sources of traffic. The traffic reflects better what really happens in a real network and consequently its results are more accurate. Our multichannel protocol tries to exploit the high capacity of the fiber by creating multiple channels on the same fiber. The advantages are the compatibility with Ethernet, the scalability (simply by adding more wavelengths to obtain higher bit rates) and the possibility of increasing the total capacity while keeping the same physical topology by increasing the bit rate per channel (thus replacing Ethernet by Fast Ethernet). CSMA/CD networks lose efficiency when the total traffic offered on the medium approaches the medium s capacity. Nevertheless, in the multichannel case the efficiency does not deteriorate so much as in the case of a single channel. Even, the average delay is much lower and the number of lost packets is a smaller by one order of magnitude when offered traffic is near the limit of the channel capacity. Compared to Fast Ethernet, our protocol maximum distance is 10 times longer, the average delay and the packet loss rate are lower, and more network capacity is easily obtained by increasing the number of channels. Our protocol offers one tenth of the offered traffic to each channel and this leads to less collisions and thus to a better throughput and average delay. Multichannel performance results are superior compared to single channel ones for the same total capacity. New optical LAN protocol exploiting WDM and coherent transmission techniques is analyzed. Conceived as Ethernet evolution our multichannel protocol shows better performance than single channel one with the same total capacity.

3 1. INTRODUCTION Forthcoming applications will need high guaranteed bandwidth and low bounded delay. This paper illustrates the advantage of distributing total capacity over several wavelengths with an appropriate protocol. Using the benefits of the multichannel capabilities, a new Ethernet 1 protocol based on a Wavelength Division Multiplexing (WDM) scheme is proposed. The major goals are to geographically extend the network up to several km, to obtain an adequate delay and throughput performances in order to support real Local Area Network traffic, to simplify the transmission and reception schemes, to be compatible with existing standards and finally to satisfy the high demand of the large number of users connected. 2. NETWORK ARCHITECTURE PROPOSAL A logical bus Local Area Network is implemented over a passive star physical topology. Each node has a tunable transmitter in order to choose the channel where to transmit its packets. It is also equipped with a coherent 3,4 receiver that captures packets from all the different wavelengths. This reception is achieved using coherent technology 5. The incoming optical signal is mixed with that of an optical local oscillator, which is shared by all the hosts, and then a group of electrical filters selects each different data channel at different intermediate frequencies. Interface to electronics LASER & modulator Electrical filters E/O STATION #1 LOCAL OSCILATOR Nx(N+1) STAR COUPLER STATION #N STATION #2 STATION #3 Fig. 2-1: Passive star physical topology with a common local oscillator and coherent reception. The essential feature of the star is that each station on the network can broadcast its packets, and it can receive information on different wavelengths (channels) independently and simultaneously from other stations. 3. PROTOCOL DESCRIPTION WDM optical networks emerge as a viable alternative to provide Gbit/s communications. A new protocol 8 based on Ethernet 1, which senses the carrier (CSMA) with collisions detection (CD) and thus which requires packet retransmission, is proposed. Channel jumping is introduced on the protocol as a new dimension. Hosts choose a wavelength to transmit data and if the chosen channel is busy, packets have the possibility to escape to another channel. The new parameters 7 are the number of consecutive collisions that one host has to wait before it changes to another wave-

4 length, and the laser tuning time, which is the time required for the laser to jump from one wavelength to another. This new protocol is compatible with Ethernet as long as those new parameters are taken from the physical layer. All simulations are done jumping randomly to another wavelength without priorities assignments to the wavelengths. All wavelengths have the same probability to be chosen. Laser tuning times are supposed to be less than an Ethernet bit time, so they do not affect on the delay measure. 4. PROTOCOL PERFORMANCE Network protocol behavior is simulated by mixing long and short fixed length packets, which is more judicious than a fixed unique packet length model. A finite number of hosts generate packets with poissonian arrivals and with a distribution of packets lengths. Then the same simulations are done with long-range dependent traffic ( 2 ). Simulation of 10 channels of 10 Mbit/s each and a single channel of 100 Mbit/s have allowed us to compare the transmission average delay (from the beginning of first attempt to transmit a packet to the end of its successful transmission). In fact, there is a standard for 100 Mbit/s Ethernet, which is called Fast Ethernet. Fig. 4-2 shows the average delay for both protocols and in both cases two different traffics sources are used ( traces and traces). As it can be seen from the picture as the network load gets higher our protocol behaves better than a single channel protocol. Comparing both protocols with sources we can see that the multichannel average delay is always below the Fast Ethernet delay. We can also notice that the sources achieve a better delay under 60 Mbit/s, but that is why we have evaluate performances with real traces which represent in a better way the real packet sources of the Ethernet network. The future real network performance will be close to the performance results. Average delay [ms] Fast Ethernet Multichannel Traffic offered [Mbit/s] Fig. 4-2: Average delay comparison between 100 Mbits and multichannel 10 Mbits for and traffic There has to be noted that the average delay does not take account on the time that packets wait in the MAC queue before being sent. We have also compute the number of packets that suffer 16 consecutive collisions, which are called as lost packets in the Ethernet network because they are resent to the upper layer (Logic Link Control, LLC). This loss measure gives an indication to understand the average delay behavior. We have represent the packet loss comparison between 100 Mbits and multichannel 10 Mbits for and traffic on fig We can see that the ratio between both protocols is on behalf of our protocol by one order of magnitude. Fast Ethernet loses 10 times more packets than our protocol does. This fact can be very important for upper layers as those lost packets effect is a signal that the network is highly loaded. The packets are resent to the network immediately and that s why the average delay is increased. Our protocol tends to spread the packets over all the channels so they have more chances to find a free interval. The minimum packet size is 512 bits (1 slot) and the maximum is 4752 bits (594 bytes). In the case, maximum length packets and minimum length packets are generated with the same probability. In the traces there are also interme-

5 diate length packets, and traffic is bursty. It is clear that both traffic sources are unequal, and thus the results can be different. It is known that Ethernet network has a different return for a different packet length, that s why we have chosen a source that generates both short and long packets. But a very important difference remains between sources and traces: stations generate packets with poissonian arrivals for the first one, in contrast packet interarrivals are bursty. The second one reflects better what happens in a real network. Packet loss [packets/s] Fast Ethernet Multichannel Traffic offered [Mbit/s] Fig. 4-3: Packet loss comparison between 100 Mbits and multichannel 10 Mbits for and traffic Throughput, which is the total number of bits per second corresponding to successfully transmitted packets, is higher in the protocol proposed, as it is shown in fig While the throughput of Fast Ethernet decreases under congestion conditions, it remains constant in the multichannel protocol. These results show that our protocol reaches to the congestion conditions for a higher traffic and it shows a better throughput. There is a straightforward connection between the packet loss and the throughput. We interpret the higher performance measures of our protocol as a result of the multichannel capabilities of the protocol. In fact, though both protocols have the same total capacity (100 Mbit/s), our protocol offers one tenth of this bit rate to each channel but not systematically to the same channel (because they change after each collision). This behavior leads to less collisions and thus to a better throughput and average delay. Throughput Multichannel Fast Ethernet Traffic offered [Mbit/s] Fig. 4-4: Throughput comparison between 100 Mbits and multichannel 10 Mbits for traffic

6 5. CONCLUSION A multichannel protocol that tries to exploit the high capacity of the fiber by creating multiple channels on the same fiber is presented. The advantages of our protocol are the compatibility with Ethernet, the scalability (simply by adding more wavelengths to obtain higher bit rates) and the possibility of increasing the total capacity while keeping the same physical topology by increasing the bit rate per channel (thus replacing Ethernet by Fast Ethernet). CSMA/CD networks lose efficiency when the total traffic offered on the medium approaches the medium s capacity. Nevertheless, in the multichannel case the efficiency does not deteriorate so much as in the case of a single channel. Even, the average delay is much lower and the number of lost packets is a smaller by one order of magnitude when offered traffic is near the limit of the channel capacity. Compared to Fast Ethernet, our protocol maximum distance is 10 times longer, the average delay and the packet loss rate are lower, and more network capacity is easily obtained by increasing the number of channels. Our protocol offers one tenth of the offered traffic to each channel and this leads to less collisions and thus to a better throughput and average delay. Multichannel performance results are superior compared to single channel ones for the same total capacity. New optical LAN protocol exploiting WDM and coherent transmission techniques is analyzed. Conceived as Ethernet evolution our multichannel protocol shows better performance than single channel one with the same total capacity. 6. REFERENCES 1. ISO/IEC ANSI/IEEE Std 802.3, Carrier Sense Multiple Access with Collision Detection (CSMA/CD) access method and physical layer specifications, IEEE publications, New York, Will E. Leland, Murad S. Taqqu, Walter Willinger, Daniel V. Wilson, On the nature of Ethernet traffic, ACM SIGComm 93, San Francisco, CA, USA, September Sadakuni Shimada, Coherent lightwave communications Technology, Capman & Hall, Milorad Cvijetic, Coherent and nonlinear lightwave communications, Artech House, Leonid Kazovsky, Sergio Benedetto, Alan Willner, Optical fiber communication systems, Artech House, G. P. Agrawal, Fiber-Optic Communication Systems, Wiley, D. Rodellar, C. Bungarzeanu, H. Garcia, C. Brisson, P.-A. Nicati, Performance Analysis of a Multiwavelength Ethernet Optical Local Area Network, Proceedings of the European Conference on Networks and Optical Communications 1997 (NOC 97), pp , Antwerpen, June D. Rodellar, C. Bungarzeanu, H. Garcia, C. Brisson, A. Küng, Ph. Robert, A new multichannel Ethernet protocol for passive optical star local area networks using coherent transmission, International Symposium on Broadband European Networks, SYBEN 98. Zürich, Switzerland. EUROPTO. May, 1998.