You are on page 1of 8

JOURNAL OF TELECOMMUNICATIONS, VOLUME 21, ISSUE 1, JULY 2013 7

Performance Evaluation of End-to-End Transmission and QoS for IP over WDM Ring Networks
Jih-Hsin Ho
Abstract—Wavelength-division multiplexing (WDM) appears to be the solution of choice for providing a faster networking infrastructure that can meet the explosive growth of the Internet. By this reason we have been devoted to develop IP over WDM ring network and accomplished many results including proposed all-optical network architecture and we have proposed carrier sense multiple access with idle detection (CSMA/ID) protocol and have done the performance evaluation of transmission algorithms for CSMA/ID WDM ring networks. We studied in depth the existing problem in CSMA/ID WDM ring networks included to achieve end-to-end transparent transmission and quality of service (QoS). We propose the address resolution mechanism between WDM ring network and heterogeneous networks. We also propose end-to-end differentiated QoS and an analytical model developed to evaluate the performance. Index Terms— IP over WDM, CSMA/ID, Heterogeneous Network Access, Differentiated QoS, Analytical Model and Simulation
————————————————

Jih-Hsin Ho is with the Department of Leisure and Information Management,Taiwan Shoufu University, Tainan 72153,Taiwan

1 INTRODUCTION

I

N the past few years, there has been an enormous increase in the bandwidth requirements of the explosion of information traffic due to the Internet, electronic commerce, computer networks, voice, data, and video, the need for a transmission medium with the bandwidth capabilities for handling such a vast amount of information is paramount. Recently, advances in solid-state and photonic technologies. The Wavelength Division Multiplexing (WDM) has been emerging as the technology of choice for increasing the transmission capacity of carrier networks. Bit rates in wavelength are 40Gbps and glass fiber is a transmission medium that permits light to travel through it without amplification for hundreds of kilometer. Furthermore, research has demonstrated that the number of wavelengths per fiber could increase to more than 1000, and this clearly is not a limitation [3]. WDM technology indeed now offers a solution for bandwidth insatiability. The backbone area network can support up to several Tbps by using WDM technology [4], so the study issue now is the access area that connects the customer premise network to Internet backbone network. Still now, the access network only use T1~T3 line. Therefore upgrading this area is very urgent problem [5]. Due to the widespread services and tremendous user population on Internet, the traffic of IP packets dominates the utilization of data networks. However, they are now transferred, switched, and manipulated through complex protocol stacks, such as IP/ATM/SONET/WDM, IP/HDLC/ SONET/WDM, and so on. How to merge and collapse the middle layers to reduce cost, complexity, and redundancy has become an important research issue [6]. Additionally, since many WDM systems have been deployed in Wide Area Networks (WANs), the bottleneck of communications will be pushed ahead from backbone networks to local access net-

works. As a result, applying WDM to LANs and MANs gains much research interests [7-10]. Although most of the research interest in WDM architectures has been focused on broadcastand-select star and wavelength routing mesh networks, there has been an increased attention focused on WDM ring networks in the past few years. Optical switching can be divided into three paradigms: Optical Circuits Switching (OCS), Optical Burst Switching (OBS), and Optical Packet Switching (OPS); OCS is a coarse switching. Because the number of the useable wavelength is constrained by the performance of the devices, wavelength routing is not the ideal switching paradigm to realize all optical networks (AON). OBS aggregated and de-aggregated a lot of packets convert asynchronous packets into synchronous burst packets and increase end-to-end delay time. Recent studies have focused on OPS and OBS and the quality of service (QoS) topics. This paper proposed a novel MAC protocol to handle the variable size IP packet completely over Metro ring and support of different QoS levels. The QoS referenced in this paper is the Differentiated Service (DiffServ) model that proposed by IETF RFC2475 and is the most suitable for backbone applications. Within DiffServ paradigm, three main node behaviors have been defined to achieve service differentiation, that are the Assured Forwarding (AF), the Expedited Forwarding (EF), and the best effort (BE) per hop behaviors. In subsequent descriptions, the structure of WDM ring network and node for the proposed protocol and its QoS scheme will be presented in Section 2. Analytical models for evaluating the average packet delay performance are developed in Section 3. To evaluate the performance of the protocol, some simulation experiments were done and their results are described in Section 4. Finally, a few remarks are given in the conclusion.

8

2. NETWORK ARCHITECTURE AND CSMA/ID MAC PROTOCOL
2.1 The Network Architecture This paper proposes a single and unidirectional fiber ring network, which connects a number of nodes. The optical fiber is composed of N data channels as shown in figure 1. The Access Points (APs) connect LAN to the MAN ring network, while PoP connects the MAN to the WAN. Each data channel makes use of one specific wavelength to convey optical signal. Therefore, based on the WDM technology, channels can work independently without mutual interference to each other. Logically, the network can be treated as a multi-ring network. The node structure of the network is shown as figure 2, each node has one tunable transmitter and N fixed receivers (i.e., TT1-FRN system) with one for each data channel. For the optical signal sent from upstream nodes, a splitter is used to tap off a small portion of the optical power detected the optical packets destination address. If the destination address in the incoming packet header matches the node address, the packet data is sent to the host. Meanwhile, the MAC controller is signaled to activate the switch (SOA) for the corresponding data channel to remove the received packet carried in the major portion of the optical signal. If the destination address is irrelevant to the node address, the detected packet is ignored and bypasses the optical packet to next node.

with Idle Detection (CSMA/ID) is proposed in [1-2] to handle variable size IP packets without complex variable optical delays or centralized algorithms. This optical MAC protocol has two features: First, it is a fully distributed, asynchronous protocol that does not need a centralized controller or a separate control channel to harmonize and synchronize the operations of nodes. Second, it supports variable length IP packets without complicated segmentation and reassembly, which becomes harder as the line speed of optical wavelengths ever increases [12].

Fig. 2. Structure of the node

Fig. 1. Structure of Metro WDM ring

Fig. 3. Cumulative distribution function (CDF) of IP packet sizes on an Internet backbone link.

2.2 CSMA/ID MAC Protocol Although the IP standard allows the length of packet between 40 bytes and 64k bytes, a measurement trace from one of MCIs backbone OC-3 links shows a discrete packet-size distribution, from 40 bytes to 1500 bytes (Figure 3.) [11]. According to the discrete IP packet-size distribution, IP packets distribute aggregated on 40, 552, and 1500 bytes. In accordance with IP packet-size distribute characteristic, we propose a node structure with packet preclassified in order to classify IP packets from the LAN into three different QoS level queues. According to the CDF (Cumulative Distribution Function) of IP packet sizes, a novel and practical Carrier Sense Multiple Access

In order to avoid packets collision and raise the network throughput. The following processes are added: The LAN network IP packets are pre-classified to three kinds of different queues (Q1, Q2, and Q3) with the buffer selector. The three kinds of queues are storage 40 bytes, 41~552bytes, and 553~1500bytes, respectively. The MAC controller is then informed by first IP packets size in the three kinds of queues. Since each node is equipped with a receiver for corresponding data channel, an IP packet can be transmitted via a corresponding available data channel to its destination node. The receiver is responsible both check the destination address of incoming IP packets and detect the available of channel to notify MAC controller.

9

According the information from (1) and (2), the MAC controller delivers a message to buffer selector that is active to transmit the Q1, Q2, or Q3 buffers packet. Figure 4 illustrates an example of the MAC controller scheme; in the example, the maximum available channel length is 552bytes and the three kinds of different queues storage first IP packet are 1200 bytes, 512 bytes, and 40 bytes respectively, and the MAC controller transmits a message to tunable transmitter which to transmit the Q2 buffers packet.

classify the corresponding queues (EF-Queue, AF-Queue, or BE-Queue) when packets from local LAN entering Metro ring. These all operations are transparent to both end-users. TABLE 1. THE QOS MAPPING TABLE

Drop Precedence EF Low

Class

LAN 101110

MetroRing 1101 0001 0010 0011 0100 0101 0110 0111 1000 1001 1010 1011 1100 0000 1110 (RSVP) 1111 (MPLS)

Class 1 Class 2 Class 3 Class

001010 010010 011010 100010 001100 010100 011100 100100 001110 010110 011110 100110 000000

Fig. 4. Example of the proposed MAC Protocol Medium

4 Class 1 Assured Forwarding Class 2 Class 3 Class 4 High Class 1 Class 2 Class 3 Class 4 Best Effort

The downstream access point can check out the incomplete IP packet by the presence of the sub-carrier signal, and pulls it off the ring. The carrier-sense can check the available channel length to notify TX for transmitting the Q1, Q2, or Q3 buffer packets. By the protocol, each node monitors wavelengths and detects available length of corresponding channel. An IP packet is transmitting onto a target channel while another IP packet is arrival on the same channel of the node; a dilemma of ring access (access collision) has occurred. The cause of access collisions is due to the node cannot predict how long the available channel will open. By the carrier access scheme, to guarantee the correctness of the protocol operations, the delay line inside nodes are used to delay the incoming packet. In addition, the delay line should be long enough to cover maximum IP packet length (1500Bytes), so the unnecessary fragmentation can be avoided, and improve the utilization of facilitate spatial on the bandwidth. Further, the fiber delay line inside the AP accounts for processing IP packets time.

Unused

2.3 QoS Support (heterogeneous) DiffServ network are proposed to promote the Quality of Service (QoS) for multimedia applications in the Internet. However, the complexity of communication between diverse local area network and the different levels QoS architectures decrease the utility of QoS provisioning. This paper presents a transparent QoS mechanism to communicate with underlying QoS architectures and also provide swappable modules to support different QoS setups for diverse applications. It integrates the local IP- DiffServ into the Metro ring, such as Figure 5; the operation is to identify the local IP-DiffServ and pre-

Table 1 illustrates DSCP format and priority class values. In order to support DiffServ service for heterogeneous LAN, this paper proposes a mapping scheme to translate DSCP values; its DSCP value as table 1, and the frame format adopted is shown in Figure 6. To find transmitted packets in optical fiber, the carrier sensing mechanism can be based on sub-carrier signaling [13] or receiver monitoring. In the subcarrier signaling, each wavelength is associated with a subcarrier frequency; when a node transmits a packet, it multi-

10

plexes the corresponding sub-carrier frequency. The nodes examine the occupancy of all wavelengths in parallel by monitoring the sub-carriers in the RF domain. The frame format is consisting of the fields as follow: SD, ED: the start delimiter and the end delimiter, they mark a physical data frame conveyed in data channels for packets. SA: source address DA: destination address DSCP MAP: mapping DiffServ DSCP values CRC: cyclic redundancy check, prevent the possible transmission errors in midway

interval between the time that the last bit of the packet reaches the destination and the moment that the last bit of the packet was transmitted. From the behavior of the expected queuing delay for the ith packet, the model can be categorized as an M/G/1 queue with vacations model [13]. Clearly, the queuing delay captures the effect of contention and is dependent on traffic density. In order to present expressions for packet transfer delay at a node on multirings using an M/G/1 vacation model, we first present some assumptions and the general notation to be used in various subsections. (1). Assumptions. For simplicity, the following assumptions are made:

1. The number of WDM channels is W. 2. The total propagation delay of the WDM ring is τ seconds, and the distances between the nodes are equal. 3. Packets which arrive are independent, identically distributed (i.i.d.) Poisson process with rate λi (packets/second) at each of the N nodes on the ring, and with aggregate arrival rate for the network of
N −1

λ =

∑λ
i =0

i

.

4. The arrival stream of packets at node i destined for node i ⊕ j is a Poisson process with rate of λi ,i ⊕ j , where ⊕
N −1

λi =

∑λ
j =1

indicates addition modulo N; thus . In the case of uniform and symmetric

i ,i ⊕ j

Fig. 5. Support DiffServ MAC Protocol

traffic on the ring, it indicates that the mean packet generation for all nodes is equal and each source sends equal traffic to all destinations.

λi = λ N , λi ,i ⊕ j =
and

λi
N −1

=

λ
N ( N − 1)

(1)

λi , i = 0 , for 0 ≤ i ≤ N − 1,1 ≤ j ≤ N − 1

Fig. 6. Data Format

5. The packets have random lengths determined at each node as independent, identically and geometrically distributed random variables (denoted by the r.v. M(bits)) with mean E[M] and probability mass function [14] Pr ( M = k ) = β ⋅ (1 − β ) k , k = 0,1,2,… 1 where β = .
1 + E[ M ]

3 PERFORMANCE EVALUATION
3.1 Approximate Analytical Models The transfer delay of a packet measured from when the packet is completely stored in the source node queue until that packet has been completely received by the destination node. This delay consists of queuing delay, transmission delay and propagation delay. The queuing delay of a packet is measured from when a packet is fully stored in a queue of the source node to the time the source node was last selected by the queue before successful transmission. Meanwhile, in this investigation, the transmission delay is defined as the interval between the source node selecting the queue to transmit the packet successfully and the time the source node last selected the queue before transmitting the packet successfully. Finally, the propagation delay of a packet is the

6. The WDM ring channel bit rate is R (bps) and the packet transmission time is X (= M/R) seconds. 7. Define MTU (maximum transfer unit) as equal to the delay line (L=1500bytes) with Ti = L/R seconds to transmit the MTU. (2). Notations. The following notations are used in the analytical formulas below: D TQi TQ αi Vi average packet transfer delay queue-waiting delay of packet i average packet queue-waiting delay residual time of packet i packet residual time duration of all the whole vacation intervals for which packet i must wait before being transmitted

11

V S

steady-state duration of all whole vacation intervals average transmission delay
Arrival time of ith packet Waiting time in queue TQi Transmission of ith packet starts

ρ Bi = ∑k =2 ∑ j = N −k +1 λi⊕k ,i⊕k ⊕ j E[ X j ]
( N − 1)( N − 2) λ E[ M ] × × 2 N ( N − 1) R ( N − 2) × λi × E[ M ] = 2× R =

N −1

N −1

(5)

Ti

P1

P2

....

....

Pi

With this assumption, the average density ρBi can be viewed as the probability of MTU is full on the ring. The probability that a packet has to wait i more MTU before it can be transmitted is ρ Bi i (1 − ρ Bi ) . The mean waiting time E[d] to find an empty MTU can be expressed as
E[d ] = ∑ i
∞ i =0

Time

L ⋅ ρ Bi L ρ Bi i (1 − ρ Bi ) = R R(1 − ρ Bi )

(6)

Busy time Residual for Mi packets time α i ahead of ith packet

Vacation time Vi due to blockage by upstream traffic

Fig. 7. Calculation of the average waiting time in M/G/1 system with vacations. The average waiting time E[TQi] of the ith packet is E[TQi] = E[αi ] + E[Mi]E[X] + E[Vi ].

The steady-state duration of all the whole vacation intervals V is equal to λiTQ⋅E[d], and combining equations (3) and (6) we obtain the average queuing delay TQ = E[α ] + λiTQE[ X ] + λiTQE[d ] (7) L ⋅ ρ Bi L = + λiTQE[ X ] + λiTQ 2⋅ R R ⋅ (1 − ρ Bi ) which can be reduced to (8)
TQ = L 2 ⋅ R ⋅ (1 − λi E[ X ] − λi L ⋅ ρ Bi ) R ⋅ (1 − ρ Bi )

6.2 Analysis of the single-ring case With the above assumptions, we model the queuing and transmission delay using a M/G/1 queue with vacations as illustrated in Figure 7. The average queuing delay, TQi, for the ith packet is given by E[TQi] = E[αi ] + E[Mi]E[X]+ E[Vi ] (2) The queuing delay and transmission delay capture the effect of contention and upstream traffic dependence. Thus we consider the delay line (or MTU) as a slot unit, so the dependence is when the full slots are uniformly and independently distributed on a single ring. Since the arrival process is assumed to be Poisson, this residual time α can be considered to be uniformly distributed between 0 and L/R. Therefore, the mean packet residual time is simply L (3) E[α ] =
2× R
→∞

Because the packet transfer delay is comprised of the queuing delay, transmission delay and propagation delay, the average packet transfer delay is D = TQ + S + τ’ (9) where τ’ is the average propagation delay from a source node to a destination node, which is often expressed as τ / 2. The average transmission delay is
S = E[ X ] + E[ d ] L ⋅ ρ Bi = E[ X ] + R ⋅ (1 − ρ Bi )

(10)

Thus, the average transfer delay is given by D = TQ + S + τ / 2

(11)

By Little’s formula, the value of limi E[Mi]E[X] is λi TQ E[X]. Letting V = lim i E[Vi ], we can thus write the steadystate version of equation (2): TQ = E[α] + λi TQ E[X]+V (4) Next we calculate approximation V by multi-channel slotted ring networks. Packets sent by an upstream source use node i as a bridge to reach their destinations, and this bridge has an average traffic load of ρ Bi = ∑ N −1∑ N −1 λi ⊕ k ,i ⊕ k ⊕ j E[ X j ].
→∞

6.3 Analysis of the multi-ring (WDM ring) case In order to analyze the multiple WDM ring networks, it is assumed distributed among the W rings, the average bridge traffic load of each ring, ρB, can be expressed as (12) ρ B = ρ Bi / W The probability that the packet at the head of a queue cannot get an empty MTU among the currently passing W MTUs is (ρB)W. Therefore, the probability that the packet has to wait i MTUs before it can be sent out is ( ρ B )W ⋅i (1 − ( ρ B )W ) . Similar to subsection III.B, let E[dB] be the average time required to find the arrival of an empty MTU, then we have ∞ L ⋅ ( ρ B )W L (13) E[d B ] = ∑ i ( ρ B )W ⋅i (1 − ( ρ B )W ) = R ⋅ (1 − ( ρ B )W ) i =0 R Since for each packet in the queue the arriving packet has to wait for L/R, the average queuing delay in the queue faced by an arriving packet is (14) TQ = E[α ] + λiTQE[ X ] + λiTQE[d B ] Therefore, we have

k =2

j = N − k +1

This upstream traffic blocks the head of the queue packet at node i. Substituting the above assumptions into ρBi gives an expression as follows:

12

E[α ] 1 − λi E[ X ] − λi E[ d B ] The average transmission delay is TQ =
S = E[ X ] + E[d B ] L ⋅ ( ρ B )W = E[ X ] + R ⋅ (1 − ( ρ B )W )

(15)

λ i,1

λ i,2

λ i,k

............
(16)

Thus, the average transfer delay is given by D = TQ + S + τ / 2

(17)

Node i

From the behavior of a priority-aware queue model, the model can be categorized as a non-preemptive priority M/G/1 queue model. Figure 8 shows model of corresponding priority k in access node i. Note that the ρi,k is the product of λi,k ( individual class k arrival rate in node i and ∑ λi , k = λi ) by E[Xi,k] ( the mean service time of class k in node i). The notation γ is the product of E[nG] by E[d]. Consider the mean waiting time Wi,1 of the highest priority queue, we obtain:
Wi ,1 = α + λi ,1 ⋅ Wi ,1 ⋅ E[ X i ,1 ] + λi ,1 ⋅ Wi ,1 ⋅ E[nG ] ⋅ E[d ]

Fig. 8. Queuing model of corresponding priority k in access node i

4. SIMULATION RESULTS
This section presents the results of the discrete event simulations under IP packet sizes on Internet backbone OC-3 link traffic. A specific network simulator has been implemented to simulate the two architectures that packets have pre-classified or no pre-classified architecture. It is important to note that all simulations were run for a time long enough to reach steadystate results. In general, it between 1 to 2 million packets was transmitted per node for each simulation. Network throughput is defined as the sum total of all traffic that enters the ring at any add/drop node and exits at another add/drop node. The simulation experiments are based on the codes by SIMSCRIPT II and are replicated corresponding to variance reduction technique with different sequences for pseudo random numbers. For simulation parameters, they are listed below Table 2. TABLE 2. THE SIMULATION PARAMETERS ARCHITECTURE (TT-FRN) NODE NUMBER CHANNEL NUMBER

(18)

α = 1 − λi ,1 ⋅ E[ X i ,1 ] − λi ,1 ⋅ E[nG ] ⋅ E[d ]
=

α 1 − ρ i ,1 − λi ,1 ⋅ γ

For the second priority queue, we have a similar expression for the mean waiting time Wi,2, except that we have to count the additional waiting time due to packets of higher priority that arrive while a packet is waiting in queue. Using the expression (18) obtained earlier, we finally have:
Wi , 2 = α + λi ,1 ⋅Wi ,1 ⋅ E[ X i ,1 ] + λi ,1 ⋅Wi ,1 ⋅ E[nG ] ⋅ E[d ] + λi , 2 ⋅Wi , 2 ⋅ E[ X i , 2 ] + λi , 2 ⋅Wi , 2 ⋅ E[nG ] ⋅ E[d ] + λi ,1 ⋅Wi , 2 ⋅ E[ X i ,1 ] + λi ,1 ⋅Wi , 2 ⋅ E[nG ] ⋅ E[d ] = = =

(19)

α + λi ,1 ⋅Wi ,1 ⋅ E[ X i ,1 ] + λi ,1 ⋅Wi ,1 ⋅ E[nG ] ⋅ E[d ]
1 − λi , 2 ⋅ E[ X i , 2 ] − λi , 2 ⋅ E[nG ] ⋅ E[d ] − λi ,1 ⋅ E[ X i ,1 ] − λi ,1 ⋅ E[nG ] ⋅ E[d ]

α + ρ i ,1 ⋅Wi ,1 + λi ,1 ⋅Wi ,1 ⋅ γ
1 − ( ρ i ,1 + ρ i , 2 ) − (λi ,1 + λi , 2 ) ⋅ γ

NO PRECLASSIFIED

PRE-CLASSIFIED TO
THREE KINDS OF QUEUES.

α
(1 − ρ i ,1 − λi ,1 ⋅ γ )(1 − ( ρ i ,1 + ρ i , 2 ) − (λi ,1 + λi , 2 ) ⋅ γ )

16 16, 8, 4 EF:20% , AF:50% , AND BE: 30%

The derivation is similar for the general priority queue (class k >1). The formula for the mean waiting time in queue is: (20) α Wi, k = ⎛ k −1 ⎡ k −1 ⎤ ⎞ ⎛ k ⎡ k ⎤ ⎞ ⎜ ⎜1 − ∑ ρi,l − ⎢∑ λi,l ⎥ ⋅ γ ⎟ ⎟ ⋅ ⎜ ⎜1 − ∑ ρi,l − ⎢∑ λi ,l ⎥ ⋅ γ ⎟ ⎟ ⎣ l =1 ⎦ ⎠ ⎝ l =1 ⎣ l =1 ⎦ ⎠ ⎝ l =1

QOS APPLYING (K = 3) LIGHT VELOCITY IN FIBER: V RING NETWORK
LENGTH

2 × 108 m / s
100 KM 10 GB/S (OC-192) 1500 BYTES (240 METERS) OC-3 TRAFFIC: 40~1500 BYTES 353 BYTES

CHANNEL SPEED SIZE OF THE
DELAY LINE

IP PACKET SIZE AVERAGE PACKETS

13

The simulation analysis for the performance rate of Access Link Load (for each node in Gbps) vs. Average Transmission-Delay (us) is shown in Figure 9. The simulation result is presented and this is a very attractive scalability feature. Figure 9 illustrates the pre-classified has fine performance compare with no pre-classified architecture, and the more increase channel number. Figure 10 to Figure 13 are illustrates simulation results for DiffServ mapping. These results are very attractive scalability feature. Regardless of the preclassified or no pre-classified architecture, EF packets have no queuing delay. AF packets have low queuing delay with the pre-classified (node number=16 and channel number=8). Even through BE packets have low queuing delay when node number equal channel number (as follow Figure 12). Above-mentioned, we proposed have a fine performance architecture of network and transparent QoS mechanism to communicate with underlying QoS architectures and also provide swappable modules to support different QoS setups for diverse applications.
μs
700
Channel Number=2 Channel Number=4

Cha nne l Num b e r=8
AF 1.00 0.90 0.80 EF BE Ave ra ge

T ra nsmission Dela y (ms)

0.70 0.60 0.50 0.40 0.30 0.20 0 1 2 3 4 5 6 7 8 9 10 Gb /s

Fig.11. IP- DifferServ with pre-classified architectur (Node Number=16 and Channel Number=8)

Cha nne l Numbe r=16

Pre-Classified

Channel Number=8
1.40

AF

EF

BE

Ave ra ge

1.20

600 500
No Pre-Classified

T ra nsmi ssi on Del a y (ms)

1.00

0.80

0.60

400 300 200 1 2 3 4 5 6 7 8 9 10

0.40 Gb/s 0.20

6

7

8

9

10

Fig. 12. IP- DifferServ with no pre-classified architectur (Node Number=16 and Channel Number=16)

Fig. 9. The pre-classified and no pre-classified architecture compare with performance
AF EF

Cha nnle Numbe r=16
BE Ave ra ge

Cha nnle Numbe r=8
1.00

AF
1.40

EF

BE

Ave ra ge

0.90 0.80

T ra nsmission Dela y (ms)

1.20

0.70 0.60 0.50 0.40 0.30 0.20 0 1 2 3 4 5 6 7 8 9 10

T ra nsmission Dela y (ms)

1.00

0.80

0.60

Gb/s

0.40 Gb /s 0.20

2

3

4

5

6

7

8

9

10

Fig.13. IP- DifferServ with pre-classified architectur (Node Number=16 and Channel Number=16)

Fig.10. IP- DifferServ with no pre-classified architectur (Node Number=16 and Channel Number=8)

14

5.

CONCLUSION

In summary, this paper proposed a novel MAC protocol for all optical WDM ring networks. The protocol supports the transmission of IP packets directly over WDM from LAN to MAN. Meanwhile, the investigation has been made about how to merge and collapse the middle layers between IP and WDM for next generation optical LANs/MANs. The CSMA/ID protocol can avoid collision packet, reuse wavelength, no fragment packet scheme and transparent QoS mechanism to communicate with underlying QoS architectures. The utilization on the bandwidth of all optical ring networks, the proposed protocol exhibits the excellent characteristics of high throughput and low delay in all optical communications. Furthermore, the pre-classification packets function of the proposed protocol have fine support QoS and priority scheme.

8th IEEE Internation Conference on Communication tems (ICCS2002), 3P-05-03, pp. 1217-1221 [13].H.M. Ghafir, Performance Analysis of a Multiple-Access Ring Network, IEEE Transaction on Communications, 41 (10) 1494-1506, Oct. 1993. [14] J. Cai, A. Fumaaglli, I. Chlamtac, “The multitoken interarrival time (MTIT) access protocol for supporting variable size packets over WDM ring network”, IEEE Journal on Selected Areas in Comm., vol.18, no.10, pp. 2094-2104, Oct. 2000. [15] C.S. Kang, A broadband ring network: multichannel optical slotted ring, Computer Network and ISDN Systems 13871398.,1995. [16] L.N. Bhuyan, Approximate analysis of single and multiple ring networks, IEEE Transaction on Computers, 38 ,1027-1040,1989. Jih-Hsin Ho received his BS degree in computer science and information engineering from Tatung University, Taipei, Taiwan, in 1993, and the MS and PhD degrees in Electrical Engineering from National Cheng Kung University, Taiwan, in 1998 and 2007, respectively. He is currently an assistant professor teaching at the Department of Leisure and Information Management, Taiwan Shoufu University, Tainan, Taiwan. His current research interests include performance evaluation, WDM networks, Internet QoS.

REFERENCES
[1] J.H.Ho, "Performance Evaluation of Transmission Algorithms for CSMA/ID WDM Ring Networks," Journal of Communications, ISSN 1796-2021, Vol.5, no.6,pp.501-508, JUNE,2010. [2] J.H.Ho, W.P. Chen, W.S. Hwang and C.K. Shieh, "Performance Evaluation of CSMA/ID MAC Protocol for IP over WDM Ring Networks,"International Journal of Communication Systems, vol.21, no.11, pp.1155-1170, Nov.2008 [3] Kartalopoulos, S.V. “Elastic bandwidth [optical-fiber communication]”, IEEE Circuits and Devices Magazine, vol. 18, pp 8 –1, Issue: 1, Jan. 2002. [4] Bissessur, H.; Charlet, G.; Idler, W.; Simonneau, C.; Borne, S.; Pierre, L.; Dischler, R.; De Barros, C.; Tran, P. “ 3.2 Tbit/s (80 /spl times/ 40 Gbit/s) phase-shaped binary transmission over 3 /spl times/ 100 km with 0.8 bit/s/Hz efficiency”, Electronics Letters, vol. 38 Issue: 8 , pp 377 -37911 April 2002. [5] Jungjoon Lee; Jun kyun Choi, “Point to point label switching protocol for optical access networks”, Electrical and Electronic Technology, 2001. TENCON. Proceedings of IEEE Region 10 International Conference on , Volume: 1 , 19-22 Aug. 2001 [6] N. Ghani, S. Dixit, T.S. Wang, “On IP-over-WDM Integration”, IEEE Comm. Magazine, vol.38, no.3, pp. 72-84, Mar. 2000. [7] J. Cai, A. Fumaaglli, I. Chlamtac, “The multitoken interarrival time (MTIT) access protocol for supporting variable size packets over WDM ring network”, IEEE Journal on Selected Areas in Comm., vol.18, no.10, pp. 2094-2104, Oct. 2000. [8] Marsan, M.A.; Bianco, A.; Leonardi, E.; Neri, F.; Toniolo, S, “An almost optimal MAC protocol for all-optical WDM multi-rings with tunable transmitters and fixed receivers”, Communications, 1997. ICC '97 Montreal, Towards the Knowledge Millennium. 1997 IEEE International Conference on , vol.1, pp. 437 –442. [9] M.A. Marsan, et al., “On the performance of topologies and access protocols for high-speed LANs and MANs”, Computer Networks and ISDN Systems 26 (1994), pp. 873-893. [10] K.V. Shrikhande, et al., “HORNET: a packet-over-WDM multiple access metropolitan area ring network”, IEEE Journal on Selected Areas in Comm., vol.18, no.10, pp. 2004-2016, Oct. 2000. [11] WAN packet size distribution. Online. Available: http://oceana.nlanr.net/NA/Learn/packetsizes.html. [12] C.C.Li, S.W.Kau, and W.S.Hwang, 2002, "A CSMA/CP MAC Protocols for IP over WDM Metropolitan Area Ring Networks,"