IJCSIS is an open access publishing venue for research in general computer science and information security. Target Audience: IT academics, university IT faculties; industry IT departments; government departments; the mobile industry and computing industry. Coverage includes: security infrastructures, network security: Internet security, content protection, cryptography, steganography and formal methods in information security; computer science, computer applications, multimedia systems, software, information systems, intelligent systems, web services, data mining, wireless communication, networking and technologies, innovation technology and management. The average paper acceptance rate for IJCSIS issues is kept at 25-30% with an aim to provide selective research work of quality in the areas of computer science and engineering. Thanks for your contributions in September 2010 issue and we are grateful to the experienced team of reviewers for providing valuable comments.
(IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, 2010 Performance Analysis of Delay in Optical Packet Switching Using Various Traffic Patterns A.Kavitha/Chettinad College of Engineering & P.Anandhakumar/Madras Institute of Technology, Technology IT Dept IT dept Chennai, Tamilnadu, India Chettinad College of Engineering & Technology, Karur, Tamilnadu, India V.Rajamni/Indra Ganesan College of Engineering, Indra Ganesan College of Engineering Trichy, Tamilnadu, India Abstract— Quality of Service parameters are improved and there is no virtually any loss, but less utilization of for development of optical packet switching technology. wavelength in this technique. In OBS, data is sent in bursts Delay is an important parameter in optical packet and a burst control message is sent ahead of each data burst switching networks and it affects the performance of the to reserve a wavelength at each hop based on the expected network. In this paper, a mathematical model is presented arrival time of the data burst. In OPS, messages are to evaluate the delay rate. Delay rates are analyzed for transmitted in packets. At each switching node, the packet fixed packet length and variable length packet for various head is processed in the electrical domain for routing traffic patterns viz. Non-uniform, Poisson and ON-OFF purpose and the packet data is kept in the optical domain. traffic models for various service classes using Wavelengths can be efficiently used in OPS . Thus the Reservation Bit technique. The results are compared with optical packet switching has emerged as one of the most the existing port based First-Fit wavelength assignment promising technologies for future telecommunication algorithm. Here delay rates are reduced by 29% in our networks. OPS utilize very high bandwidth in the optical class based model than the port based model. fiber using WDM. WDM offers an aggregate throughput of Keywords-component; Optical Packet Switching (OPS), the order of terabits per second. WDM is widely becoming RB (Reservation Bit algorithm), FF (First-Fit Wavelength accepted as a technology for meeting growing bandwidth assignment algorithm), Quality of Service (QoS), Packet demands, and WDM systems are beginning to be deployed Loss Rate (PLR), BER (Bit Error Rate), WDM in both terrestrial and undersea communication links. Thus, (Wavelength Division Multiplexing). WDM offers an excellent platform for carrying IP traffic. Consequently, OPS technology has many advantages, attract I Introduction more intensive attention than ever. The next generation telecom infrastructure definitely comprise of optical Due to the explosive growth of internet applications in networks with improved QoS. recent years, data traffic has been exceeded the telephony In order to improve the performance of QoS in optical traffic and bandwidth demands have been continuously packet switching network, a detailed study has been made in increasing. It is also expected of the future networks to this paper. The performance analysis of optical packet transport heterogeneous traffic services including switching consists of two important issues namely packet multimedia and interactive applications necessitating loss and delay. Inorder to provide better QoS in optical bandwidth guarantees, minimum delay, less PLR, controlled packet switching, PLR and delay should be reduced. Packet jitter and etc. QoS provisioning seems therefore a mandatory loss and delay are not new issues in optical networks; task. Optical networks offer an extremely high traffic however minimum loss and delay provide better QoS in bandwidth capable of providing communication channels Optical packet switching. In our earlier report the for several hundred nodes. Thus, the network traffic requires performance of 8B/10B code, Systematic code and Viterbi the network to evolve by increasing transmission capacity of code in optical transmission in terms of Bit Error Rate has optical fibers as well as switching capability. There are three been analyzed. In the physical layer, transmission is done on switching schemes in optical networks namely optical bit by bit basis and Bit Error Rate has been reduced in the circuit switching, optical burst switching (OBS) and optical physical layer. When the bit errors in the physical layer are packet switching (OPS). In the optical circuit switching, a rectified, it will reduce the packet loss rates in the higher dedicated end-to-end light path is established for each layers [2,3]. We have already reported that PLR has been connection. Thus the transmission delay can be guaranteed 238 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, 2010 reduced in non-uniform traffic pattern . We have also implemented a RB algorithm for minimizing PLR in buffered non-uniform traffic pattern of optical packet switching mechanism . In this paper we have analyzed the asynchronous OPS in terms of delay. Delay rate for fixed length packet and variable packet length has been studied for various traffic patterns viz. Non-uniform, Poisson and ON-OFF traffic models. This paper has been organized as follows: Section II describes the architecture of asynchronous OPS. In section III, the analysis of delay rate is carried out for various traffic patterns in asynchronous OPS. In section IV, results along with discussions are presented and Section V deals with the conclusion and future work. Fig 1. Model of the switch in OPS network under consideration II. Description of Architecture The architecture used in this paper is presented in [4,5]. It The following assumptions are made for the simulation: has been reproduced for the reference. The size of switch under consideration is N X N. The switch has F input fibers Let ji denote the number of class i packets that arrive in a and F output fibers. By utilizing WDM, each fiber provides time slot. The total number of packet arrivals at the fiber in a N wavelengths to transport data with a capacity of C bps. time slot is k. That is j0 + j1 +… +jd-1 = k. In order to isolate Buffers with the size of 5 are used in the OPS switch the service classes, the parameter li (0 ≤ li ≤ N) is catering to each of the service classes. The switching introduced, which is the number of wavelengths reserved for process in OPS can take one of the two main forms. It can Class i traffic in case of contention in a time slot. For a be synchronous (time slotted) with the fixed packet length or service class i, if ji < li (the number of incoming packets are asynchronous (non-slotted) with variable packet lengths. In less than the wavelengths assigned), it will result in ji - li synchronous operation mode, all arriving packets have a free slots. For a service class i, if ji > li (the number of fixed size and their arrival on each wavelength is incoming packets are greater than the wavelengths synchronized on a time-slot basis, where a time slot is the assigned), resulting in li - ji overflow of packets. time needed to transmit a single packet. The RB algorithm  is implemented by introducing The operations of optical packet switching can be briefly buffers. The RB algorithm is used for slotted OPS wherein described as follows: When a packet arrives at the switch, buffers are used to avoid overlapping of packets. This the packet header is extracted and processed electronically algorithm is designed and used to improve the QoS of the by the control module. While the header is processed, the networks in terms of reduction in PLR. By using the same packet payload is buffered in the optical domain using FDL algorithm delay rate is found and analyzed in slotted OPS processing buffers. Based on the destination, information with fixed packet lengths for various traffic patterns. Hence extracted from the packet header and the control module QoS is improved in terms of reduction in PLR and delay decides to which output fiber/wavelength the packet is [4,5]. In this paper the same algorithm is implemented in switched and configures the switch accordingly. Contention asynchronous OPS for the packets of variable lengths. occurs when two or more packets are assigned to the same output port on the same wavelength at the same time [6 and III. Operating Principle 7]. The network has d service classes, ranging from service Operation of optical packet switch in a synchronous manner class 0 to service class d-1. We assumed that the output on a with fixed length of packets is explained in . Here we single fiber/wavelength as the tagged fiber/wavelength. assume that optical packet switch operates in an Delay is calculated for class i traffic at the tagged output asynchronous manner with variable length of packets. fiber. Fig1 shows a switch for packet arrivals to a tagged Analysis of delay rate for variable packet sizes in output fiber must originate from one of the FN input asynchronous OPS using three types of traffic patterns viz. wavelengths. Non-uniform, Poisson and ON-OFF traffic models is presented in this paper. In contrast to , immaterial of packet size, allotted time slots remain the same and no shifting of time slot allocation is done. In non-uniform traffic pattern all nodes are not to receive and send similar volumes of traffic [4, 9]. The number of packets arrived and transmitted of packets is not equal. There is an incoming packet for every slot. We assume that 239 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, 2010 each packet has equal probability of 1/N being addressed to this paper, at every node there should be arrival(s) of any given output port and successive packets that arrive at different classes of packets. In ON-OFF model, ON periods the tagged output/wavelength are independent. Also, the and OFF periods occur at each node for different classes. packet arrival time at the input and transmission time to the When one class is in ON period, other classes may be in output port are not equal. This is due to randomness of OFF period at that particular node and OFF period of one packet arrival. This increases the occupancy of the free class is used by the other classes for transmitting packets. slots. In Poisson traffic model, packet arrivals are random Also, other possibilities are (1) all classes may be in OFF and are mutually independent. The Poisson distribution can period, (2) all classes may be in ON period and (3) some of be obtained for packet arrivals during an infinitesimal short the classes may be in ON period and rest may be in OFF period of time λ t, where λ is called the arrival rate. Packet period. arrival at the output fiber is a single Poisson process with λ. We assume that the switch is having buffers. In RB In Poisson model, the number of arrivals in non-overlapping algorithm, the packets use the wavelength according to their intervals is statistically independent. The arrival rate λ is service classes. If there is flow of packets into the ports expressed as the average number of arrivals during a unit of with a specified service class, the incoming packets with time. The time distance between consecutive packet arrivals assigned wavelengths occupy the ports as per the assigned is exponentially distributed . However Poisson arrival service class. When the assigned wavelengths in one service model is not an accurate model considering packet arrivals class are occupied, it checks for the free wavelengths of at the tagged output fiber/wavelength in OPS. The Poisson other service classes and the packets will occupy the free arrival model assumes an infinite number of sources, which slots in the other service classes. When the assigned is not the case in a real switch. When using a Poisson arrival wavelengths are completely occupied in all the ports, the model, there is the possibility of independency between the packets overflow. By introducing buffers, the packets that service time and the packet inter-arrival time . overflow are saved. When a free slot is not available for an Bursty traffic is also known as ON-OFF traffic model. Both incoming packet, instead of dropping that packet, it will be the ON and OFF periods are distributed using exponential saved in the buffer which is provided in the switch. Before distribution with the mean 1 μ and 1 λ respectively. It is entering the buffer, a bit is added in the packet header for the purpose of reservation with respect to their service an alternating process where an OFF period follows an ON classes. Whenever free slots are available, the packets in the period, and an ON period follows an OFF period. During buffer occupy free slots. Buffered packets will have the ON periods, series of packets are transmitted from the priority over the incoming packets. In the FF algorithm, all source node and the time is called active periods. OFF wavelengths are numbered in a certain order, for example periods are called passive periods and no packets are ascending order from 0 to W-1, where W is a number of transmitted from the source. Active periods of a source are wavelengths. When the deciding port attempts to assign a exponentially distributed with one specific mean value, and wavelength, it sequentially searches all wavelengths in an passive periods are exponentially distributed with another ascending order and assigns the first available wavelength mean value. During an active period, packets are generated . In class based model, each node transmits the packets at regular periods. The most commonly used VOIP traffic according to their classes. Buffers are placed in the port for model is based on a two-state on-off model of a single voice every class. In port based model, wavelengths are placed in source. When a voice source is transmitted, it is in the ON a sequential order. Irrespective of the class, the available state, and when the voice source is silent, it is in the OFF wavelength is used by the incoming packets in a sequential state; and the ON and OFF states appear alternatively. ON- order and buffers are only placed in each port. OFF sources, each of which exhibits a phenomenon called FF algorithm is implemented in port based and packets are the “Noah Effect” resulting in self similar aggregate traffic. transmitted according to their wavelengths, whereas The Noah Effect for an individual ON-OFF source model transmission of packets is class based in RB algorithm and results in ON and OFF periods, i.e. “train lengths” and packets are transmitted according to their service classes and “intertrain distances” that can be very large without wavelengths. Thus the drop rates of packets are reduced in negligible probability [12 and 13]. The number of packet optical networks resulting in improved QoS. The delay rate arrivals divides the timeslot and produces the time slice is found and analyzed in asynchronous OPS for various which is utilized for packet transmission. Hence the packet traffic patterns viz. Non-uniform, Poisson and ON-OFF arrival rate is random, time slice is determined according to traffic models for various service classes with packets of randomness of the packet arrival. ON and OFF periods are variable lengths. The delay rates for RB algorithm and FF used with their own mean values and using the same mean algorithm are found and compared. values, various packets arrive and are transmitted during ON For fixed size packets and variable size packets in OPS, period and idling occurs during the OFF period. During ON delay rate is encounted by implementing RB algorithm. The period, packet is transmitted and there is no flow of packet packet includes payload and header. Fig 2 shows the packet during OFF period. In this model, FN independent state header. For fixed size packets, size considered is 512 bytes. model generate packet arrivals at a tagged output fiber. In For variable packet size, the range of packet size is in the 240 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, 2010 range of 512 bytes and 2k bytes. 20 bytes is included as slot. Buffers are used along with RB technique, 240 packets header along with the payload invariably for any type of are chosen with an Erlang load of 1.5. packets. Variable size packets are used in VoIP applications. In variable packet size, the size has been controlled by the application. Packet sizes of the range of 1024 to 2048 bytes show good efficiency in terms of bandwidth and reliability in Digital Video Broadcasting . Packet size is measured using uniform min and max distribution. Packet size is chosen depending on the application. Source Desti Source Desti Packet Time Flow IP -nation Port -nation Sequence Stamp id address IP Port no address 4 4 2 2 2 4 2 bytes bytes bytes bytes bytes bytes bytes Fig. 2 Structure of the packet header When a packet is send from one node to other, the following delays occur: (1) transmission delay (time required to send Fig 3. Delay rate for various traffic patterns for service class 3 using reservation bit technique. all bits of packet into the wire), (2) propagation delay (the time taken by the packet to travel through the wire), (3) processing delay (the time taken to handle the packet in the network system), and (4) queuing delay (the time taken is buffering the packet before it can be sent). In most cases, the delay (2) and (4) are considered in simulations and measurements. The transmission delay (1) is usually small for fast links and small packets and is therefore not considered. Traditionally, the processing delay (3) has also been negligible . In our measurements, we consider the delay (2) and (4). The mean delay for the above said traffic pattern is found using equation (1). TD = TPr opagation + TQueue (1) Fi Initially, we calculate the propagation delay that occurs g 4. Delay rate for various traffic patterns for service class 3 using First- when a packet travels from the source to the destination. Fit wavelength assignment algorithm Next, the queue delay experienced by a packet is calculated using equation (2). This delay is due to waiting period of the packet in the queue. A packet is in a queue, if a free wavelength is not available at that particular time slot. 1 TQueue = N ∑ (T ) i =0 i (2) where Ti is the transmission time of class i packets at particular time slice. Summation of the waiting time in the buffer and the transmission time between source node and destination node through the switch is considered as delay and the same is found for the above said traffic patterns. IV. Results and Discussions The delay values for the fixed length packets in slotted Fig 5. Delay rate for various traffic patterns for service class 4 using OPS using RB algorithm is studied and is also compared to reservation bit technique the FF algorithm in our earlier paper . In this paper, a detailed analysis is carried out to find delay in asynchronous OPS for variable length packet and is compared to FF algorithm. We consider 240 packets for the simulation purpose. Delay for class i packets are calculated in the tagged fiber. Wavelength assigned is 16 and total time slot chosen is 10, hence this architecture can transmit 160 packets in a time 241 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, 2010 Fig 6. Delay rate for various traffic patterns for service class 4 using First-Fit wavelength assignment algorithm Fig 9. Delay rate comparison Fig 7. Delay rate for various traffic patterns for service class 5 using reservation bit technique Fig 10. Comparison of data transmission In service class 3, RB technique has 15 buffers, each service class has 5 buffers, but in FF algorithm, each port has 5 buffers and the total number of buffers is 20. 10.96ms, 10.88ms and 9.83ms are the delay values while employing Fi FF algorithm and 7.55ms, 7.79ms and 7.91ms are the delay g 8. Delay rate for various traffic patterns for service class 5 using First- values while employing RB technique in asynchronous OPS Fit wavelength assignment algorithm using Non-Uniform, ON-OFF and Poisson traffic model respectively for Service class 3 and the same is shown in figs 3 and 4. In Service class 4, both the techniques have 20 buffers. 10.94ms, 10.56ms and 10.18ms are the delay values while employing FF algorithm and 7.77ms, 8.01ms and 7.561ms are the delay values while employing RB technique in asynchronous OPS using Non-Uniform, ON-OFF and Poisson traffic model respectively for Service class 4 and 242 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, 2010 the same is shown in figs 5 and 6. From these it is seen that uniform traffic pattern result in better Quality of service both the techniques have same amount of buffer. FF compared to ON-OFF model and Poisson model. At the algorithm exhibits more delay, but delay is less in our same time ON-OFF traffic pattern have better QoS when approach. compared to Poisson process if OFF periods of one service While in service class 5, RB techniques have 25 buffers and classes are more efficiently utilized by other service classes. FF algorithm need 20 buffers. 10.09ms, 10.88ms and  Biao Chen and Jianping , “ Hybrid Switching and P-Routing for Optical Burst Switching Networks”, IEEE Journal on Selected areas in 9.50ms are the delay values while employing FF algorithm communications, vol 21, no 7,pp.1071-1080, Sep 2003. and 7.1ms, 6.27ms and 6.48ms are the delay values while  A.Kavitha, V.Rajamani and P.AnandhaKumar, “Performance employing RB technique in asynchronous OPS using Non- analysis of Coding Techniques to find BER in Optical Transmission”, Uniform, ON-OFF and Poisson traffic model respectively IEEE 1st International Conference on Advanced Computing- ICAC 2009, 13-15 Dec. 2009 Page(s):21 - 27 for Service class 5 and the same is shown in figs 7 and 8.  A.Kavitha, V.Rajamani and P.AnandhaKumar, “Evaluation of BER For all the service classes under consideration, buffers are in Optical Packet Switching using Various Coding Schemes” IEEE more or less the same for RB technique and FF algorithm, Transaction on Optical Communication and Networking (under review). whereas the delay rate is slightly higher in FF algorithm.  A.Kavitha, V.Rajamani, “Performance Analysis of Slotted Optical Switching Scheme using Non-Uniform traffic” – Journal of Optical The above statement is true when the class based Communications, Vol. 29, July 2008, pp.107-111. transmission is compared with port based transmission.  A.Kavitha, V.Rajamani and P.AnandhaKumar, ”Performance Simulation results show that for all service classes under Analysis of Slotted Optical Packet Switching Scheme in Non-Uniform any type of traffic pattern, class based model produces 29% Traffic Pattern Using Reservation Bit Technique”- Selected for publication- INFOCOMP, Journal of Computer Science. reduction of delay rate when compared to port based model  H. Øverby and N. Stol, “Evaluating and comparing two different and the same is shown in fig 9. In RB technique the buffered service differentiation methods for OPS: the wavelength allocation packets with respect to their classes will occupy the free algorithm and the preemptive drop policy,” in Proc. 3rd Int. Conf. slots of the corresponding service class wavelengths on First Networking, vol. 1, Feb. 2004, pp. 8-15.  S.Bjornstad, N.Stol and D.R.Hjelme, “A highly efficient optical Come First Serve basis which reduces the waiting time in packet switching node design supporting guaranteed service” in proc. Of the buffer whereas in FF algorithm wavelength utilization is SPIE, Vol 4910, 2002, pp.63-74. sequential, so buffered packets wait until they are serviced  BO Wen, Ramakrishna Shenai, and Krishna Sivalingam,” Routing with the next sequential order of wavelengths. Thus delay Wavelength and Time-Slot-Assignment: Algorithms for Wavelength- Routed Optical WDM/TDM Networks”, Journal of Lightwave rate is less in asynchronous OPS when RB technique is Technology, vol 23, no 9, pp.2598-2609,Sep 2005. employed. The delays are improved much in our algorithm.  Adisak Mekkittikul, Nick McKeown “Scheduling VOQ Switches This is due to the class based transmission. under Non-Uniform Traffic”, CSL Technical Report, CSL-TR 97-747, Fig 10 shows the total number of transmitted bytes per slot. Stanford University, 1997 Stanford University,Standford.  Wuyi Yue, Yutaka takahashi,hideaki tagaki, “Advances in It is shown that more number of packets is transmitted in Queueing theory and network application:, Springer,ISBN: 978-0-387- class based model for all service classes for all traffic 09702-2, e-ISBN: 978-0-387-09703-9. patterns under consideration when compared to port based  Harald Overby and N.Stol “ Quality of Service in synchronous model. Hence class based model produces lesser delay as bufferless optical packet switched networks” Kluwer Telecomm. Sys. Vol 27, 2004. pp. 151- 179. well as more number of bytes transmission when RB Mihails Kulikos and Ernests Petersons”Remarks Regarding technique is employed. It is also shown that non-uniform Queueing Model and Packet Loss Probability for the Traffic with self – traffic pattern produces lesser delay when compared to ON- Similar Characteristics”, Networks”, International Journal of Computer OFF and Poisson model. Science 3;2,Spring 2008,pp. 85-90.  Eric W.M. Wong and Moshe Zukerman,” Bandwidth and Buffer Tradeoffs in Optical Packet Switching”, Journal of Lightwave V. Conclusion Technology, vol 24, no 12, Dec 2006, pp 4790- 4798. The delay rates for the Non-uniform, Poisson and ON-OFF  Xuehong Sun, Yuunhao Li, Ioannis Lambadaris and Yiqiang traffic models for various service classes are analyzed. Q.Zhao, Performmance Analysis of First – Fit Wavelength Assignment Algorithm in Optical Networks”, IEEE 7th International Conference on Minimum delay has been achieved in all service classes Telecommunications – ConTEL2003, June11-13, 2003, pp. 403-409. under consideration by using reservation bit technique. It Vadakital, V.K.M. Hannuksela, M.M. Razaei, M. Gabbouj, provides lesser delay for all service classes for fixed and M,” Optimal IP Packet Size for Efficient Data Transmission in DVB-H”, variable length packets when compared to first-fit Proceedings of the 7th Nordic ,IEEE Signal Processing Symposium, pp. 82-85, 7-9 June 2006. wavelength assignment algorithm. It is seen from our  Ramaswamy Ramaswamy, Ning Weng and Tilman Wolf, simulation, Poisson arrival model which is assumed in the “Characterizing Network Processing Delay”, Proc. in Globecom 2004, analysis approximates a more realistic model wherein all IEEE Communication Society, pp.1629- 1634. input wavelengths are modelled as independent on/off processes with exponential holding time. Also we have presented a comparative study of reservation bit technique AUTHORS PROFILE and first-fit wavelength assignment algorithm for synchronous and asynchronous OPS. It is concluded that in Optical packet switching reservation bit technique reduces delay; hence QoS is improved and at the same time Non- 243 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, 2010 A.Kavitha received the B.E degree in Electronics and Communication Engineering from Madurai Kamaraj University, India in 1997. The Master of Engineering degree in Computer and Communication Engineering from Anna University, Chennai in 2004 and currently pursuing Ph.D degree in Optical Communication Networks in Anna University, Chennai. At present she is working as a Senior Lecturer in Chettinad College of Engineering and Technology, Karur. She has published more than 10 papers in referred National and International conferences/Journals. Her area of interest includes Networks, Computer Architecture, Digital Communication and etc. V.Rajamani received the B.E degree in Electronics and Communication Engineering from National Engineering College, Anna University, India in 1990. The Master of Engineering in Applied Electronics from Government College of Technology, Bharathiyar University Coimbatore in the year 1995 and the Ph.D. degree in Electronics Engineering from Institute of Technology, Banaras Hindu University, Varanasi in 1999. He started his carrier as Lecturer in Mohamed Sathak Engineering College from 1991 onwards. He has held various positions in various Engineering Colleges. At present he is working as a Principal in Indra Ganesan College of Engineering, Tiruchirappalli. He has completed a project under AICTE - RPS scheme successfully. He has published more than 80 papers in referred National and International Journals/ conferences. His area of interest includes Device modeling, VLSI, Image processing, Optical Communication and system. P.Anandhakumar received the B.E degree in Electronics and Communication Engineering from University of Madras, India in 1994. The Master of Engineering degree in Computer Science and Engineering from Bharathiyar University in 1997 and Ph.D degree in Computer Science and Engg. from Anna University, Chennai in the year 2006.At present he is working as a Assistant Professor in Madras Institute of Technology, Chennai. He published more than 50 papers in referred National and International conferences/Journals. His area of interest includes Digital Communication, Soft Computing, Robotics and etc. 244 http://sites.google.com/site/ijcsis/ ISSN 1947-5500
Pages to are hidden for
"Performance Analysis of Delay in Optical Packet Switching Using Various Traffic Patterns"Please download to view full document