Enhancing Bandwidth Utilization and QoS in Optical Burst Switched High-Speed Network

91 Enhancing Bandwidth Utilization and QoS in Optical Burst Switched High-Speed Network Amit Kumar Garg and R S Kaler School of Electronics and Communication Eng, Shri Mata Vaishno Devi University (J&K), India garg_amit03@yahoo.co.in Abstract In this paper, an efficient integrated novel scheme has been proposed that reduces the probability of burst contention by controlling the route at an edge router without contention resolution scheme at a core router. Also, this paper has been focused on the number of resources a packet uses in the network to lower the overall resource usage and loss probability. Simulation results indicate that the proposed scheme enhances both end-to-end throughput and Quality of Service (QoS).By leveraging statistical multiplexing, re-configuration is minimized and bandwidth utilization is enhanced. The blocking probability of the proposed scheme is also significantly reduced as compared with the general OBS. Index Terms Optical Burst Switching, Priority Scheduling Based on Hop- Count (PSBHC), Wavelength Reservation Algorithm. I. INTRODUCTION Increasing demands for transmission bandwidth driven by the growth of IP (Internet Protocol) based data traffic, especially real time multimedia services, give rise to Dense Wavelength Division Multiplexing (DWDM) technology which make possible to exploit the huge potential bandwidth of optical fibers. Though Optical Packet Switching (OPS) technology can be attractive for all optical backbone networks, this technology has some technological limitations such as Optical RAM and all optical processing. Presently, Optical Burst Switching (OBS) technology [1-2] is under study as a promising solution for optical Internet backbone in the near future, since OBS eliminate the electronic bottleneck at switching node with the help of no O/E/O conversion and guarantee the Class of Service (CoS) without any buffering. The OBS network can be envisioned as two coupled overlay networks: a pure optical network transferring data bursts and a hybrid control network transferring control header packets. The control network is just a packet switched network, which controls the routing and scheduling of data bursts in the all optical network based on the information carried in their control header packets. This coupled overlay networks take advantage of both mature electronic control technologies and promising optical transport technologies. Many approaches and architectures have been proposed in literatures to carry information in optical domain. Among them, optical burst switching (OBS) and wavelength routed network seems to be the most successful. In [3] a new novel architecture has been proposed that uses both methods in order to overcome the limitations imposed by each approach. The proposed architecture deployed bursty

92 traffic in a hybrid fashion where implicitly predicted and explicitly pre-booked traffic were dynamically allocated reserved end-to-end paths, inheriting the spirit of conventional wavelength routing; whilst, the non-predicted traffic was transmitted via classical OBS reservation mechanism(s) with the best efforts support. The complete network structure along with load-balancing prior reservation strategy was presented. Simulation results revealed the performance of the proposed work by examining the blocking probability and delay characteristics. The encouraging results encouraged stimulation for further work on optimal traffic placement, QoS provisioning and various a priori resource reservation strategies. Since Optical Packet Switching (OPS) and most OBS schemes send data without waiting for any setup acknowledgement, blocking may occur. Blocking is resolved in the wavelength domain, time domain or space domain. Optical buffering and wavelength conversion are technologies that can be used in a core router to suppress contention. However, these technologies are not still in practical use because of the issues of the performance and cost implications. Optical Burst Switching (OBS) aims to provide higher utilization and flexibility than the current optical paradigms of multiple opticalelectronic-optical conversions. One of the critical design issues in OBS is finding ways to minimize burst dropping resulting from resource contention. In [4], a new integrated scheme based on resourcereservation and adaptive network flow routing to alleviate contention in optical burst switching networks, was proposed. The proposed scheme reduced the overall burst loss in the network and at the same time avoid the packet out-of-sequence arrival problem. Simulations were carried out to assess the feasibility of the proposed scheme. Its performance was compared with that of contention resolution schemes based on conventional routing. Through extensive simulations, it was shown that the proposed scheme not only provides significantly better burst loss performance than the basic equal proportion and hop-length based traffic routing algorithms, but is also void of any packet re-orderings. The attractiveness of this approach lies in the preservation of packet ordering of individual flows while reducing the overall burst loss in the network. Till now, no work has been done to reduce burst contention by controlling the route at an edge router without contention resolution scheme at a core router. In this paper, an efficient integrated novel scheme has been proposed that reduces the probability of burst contention by controlling the route at an edge router without contention resolution scheme at a core router. In a conventional OBS network, a burst is forwarded on the shortest path route. In the proposed scheme, each edge router selects a suitable route to the destination edge router autonomously by using feedback mechanism and prior-information packets. Moreover, this paper is also focused on the number of resources a packet uses in the network to lower the overall resource usage and loss probability.

93 II. MODEL O OPTICAL BURST SWITCHING NETWORK OBS combines the benefits of optical circuit switching and optical packet switching [2] and as such, it maintains the efficiency of optical packet switching while reducing the implementation complexities. It takes advantage of the buffering and processing capabilities of electronics at network edge, combined with benefit of optical bypass and wavelength routing in the core. A. Components of OBS Edge Router The key components of OBS edge router are: Packet Assembler Unit, Data Channel Unit and Control Unit (as shown in ig.1). Packet Assembler Unit (PAU): It collects the input packets and places them in certain burst boxes to form bursts. The incoming buffer acts as an assembler for data flows. The scheduler also makes burst request to control unit to initiate a control information packet. Statistical multiplexing is introduced to increase utilization of bursts or data channels. Data Channel Unit (DCU): It is the abstraction for a potential wavelength channel to transmit a burst. There are two states for a data channel. In the reservation state, resources such as wavelength, transceiver and OXC ports are assigned to the data channel and hence a wavelength channel is setup. After the burst is transmitted, the wavelength channel is torn down and the resources are released. In the release state, the data channel just waits for new reservation command from control unit. Control Unit (CU): The control unit takes care of the control packet. The control packet should include enough information for burst routing, offset time (on which the burst will arrive) and burst length. Buffers Scheduler Bursts Data Channel Unit E/O E/O Packet Assembler Unit Control Packet Processing/Transmission Control Unit ig.1. Components of OBS edge router

94 B. Time Relations among PAU, DCU & CU or a data channel to transmit a burst, four parts compose a reservation of a data channel, i.e, the time for a reservation state is at least: T reservation (Min ) = T setup + T burst + T prop + T teardown (1) Where T reservation and T teardown are time for physical actions, if any, respectively to setup and to teardown of a wavelength channel after receiving the allocation or de-allocation request from the control unit. The T prop is the propagation time the burst takes to pass the channel. T burst is the transmission time for the burst at the full speed of the wavelength channel. The three parts except for the T burst are a cost incurred by re-configuration. If re-configuration takes place frequently, these parts could be accumulated to be a considerate part and lessen the bandwidth utilization. Therefore, statistical multiplexing is used to reduce the chance of re-configuration, in which case the time for setup and teardown is zero and even T prop could be excluded in computing the minimum Treservation. The re-configuration bandwidth cost ( ) reconfiguration C and bandwidth utilization ( ) U can be defined as: C reconfiguration T reservation burst = (2) T T reservation burst U = (3) T T reservation Using (2) & (3), it implies that Creconfigua tion = 1 U (4) A burst data will not be transmitted before the resource reservation of the previous burst occupied is released. Thus an average end-to-end delay for a burst is: T + delay = T reservation T release (5) Where T release is the release time or idle time for a data channel. In equation (5) the average is made over all used data channels. Again, in the case of statistical multiplexing of flows sharing same routing, re-configuration is avoided and delay is identical to all these flows. III. PROPOSED INTEGRATED SCHEME A. Prioritized Scheduling Based on Hop-Counts (PSBHC) The proposed prioritized scheduling scheme is based on hop-counts, which considers several parameters, such as the packet s time in the network, the application s real-time demand, the number of resources used in the network and the numbers of resources left to use on its path in the network. The JET scheme experiences increasing blocking probability for each hop performed (i.e. resource used). This has the undesirable effect of wasting resources and increasing the network s overall loss probability. The proposed scheme not only utilizes the network resources effectively, but also treats the packets fairly. Each edge router keeps the information of all routes to each destination edge

95 router. The priority is set for each route. The source edge router receives a feedback packet after sending a burst. When a burst is forwarded successfully, the destination edge router sends back the feedback packet that indicates the success of the transmission, whereas when a burst is discarded at an immediate core router, the core router sends back the feedback packet that indicates the failure of the transmission. Each route has two values P and N. P is the priority ( 0 1) number of received feedback packets. The default value of P is1, the default value of P and N is the N is also1. On receiving feedback packets, edge routers update P and N by using formulas [2] written below, P N + 1 Success : P =, N = N + 1 (6) N + 1 P N ailure : P =, N = N + 1 (7) N + 1 When the priority P is close to1, the route has a high probability of the success. When the priority P is close to 0, the route has a high probability of the failure. In order to improve the performance; the source edge router sends not only a burst but also some prior-information packets on the control channel. Prior-information packets are forwarded on several routes except the route used for the transmission of a burst. By sending these packets, an edge router searches whether the transmission of a burst is succeeded or failed on the route, except the route used for the transmission of the burst. The transmission of these packets enables an edge router to find a suitable route without sending a burst payload and discarded burst number can also be reduced. B. Bandwidth Reservation Algorithm In the proposed scheme, two classes of services, real-time and non-real-time, have been considered. The bursts in the real-time class have a strict bound on delay and delay-jitter, thus requiring a guaranteed low blocking probability. On the other hand, the bursts in the non-real-time class can tolerate delay but require reliable delivery which can be accomplished by buffering and retransmissions [5-6]. In this paper, it has been assumed that no buffers are used in the optical layer, which is highly desirable in all-optical networks. However, buffering is used in the electrical layer for control packets in OBS nodes. The control packet [6] contains the information about its corresponding burst and is electronically processed by the ingress OBS node and at all the subsequent nodes along the path to the destination user. Therefore, the control packets cannot be transported transparently in an OBS network. It is feasible to buffer the control packets in the electrical layer at the OBS nodes. Two queues are added in JET protocol, q 0 for class 0 control packets and q 1 for class 1 control packets. The size of q 0 and q 1 is limited by the memory resource in the OBS node. Moreover, a time window Δ t is associated with these two queues. The control packets will be in a particular window[ t, 1 t1 + Δt], if the control packet arrives in that interval. All incoming control packets in the particular window are

96 buffered and are kept in the corresponding queues. There is offset-time between the control packet and its corresponding burst. A bandwidth reservation algorithm is invoked to allocate bandwidth for the bursts in the two queues. urthermore, irst Come irst Served (CS) scheme is used for control packets in the same queue. The bandwidth reservation algorithm at a single node is shown in ig.2.the bandwidth reservation algorithm shows that class 1 is always scheduled before class 0 in time window Δ t, which indicates that class 1 has more priority than class 0 in reserving bandwidth. Buffer all the control packets in time window Δt to the corresponding queues or each control packet c j in q 1 If resources are available for c j Reserve bandwidth forc j ; Else Blockc j ; End or or each control packet c j in q 0 If resources are available for c j Reserve bandwidth forc j ; Else Blockc j ; End or ig.2. Bandwidth Reservation Algorithm IV. IV. SIMULATIONS AND RESULTS The performance of the proposed scheme has been evaluated using NS-2 simulator [7] on the NS14-Nodes network (as shown in ig.3). A. Assumptions A bidirectional link consists of two unidirectional fibers in opposite direction. Each fiber has 4 data channels at 1 Gb/s transmission capacity. In order to get more realistic results, the long range dependent traffic model is employed. In this traffic model, traffic that arrives at each node pair in the network is the aggregation of multiple IP flows. Each IP flow is an ON/O process with Pareto distributed ON and O times. The packet length is set to be 100 bytes. There is no wavelength converter and optical buffer in all core routers. Also, blocking probability includes source and ρ destination busy conditions. The offered load ratio of class 1 to class 0 be set at1 : 9. is α defined as the offered load of the other trafficα 1, α is called the traffic bias.

97 1 11 12 2 4 5 7 8 9 14 3 13 6 10 ig.3. NSNET with 14 Nodes B. Simulation Parameters Wavelengths= 3-12 per fiber Average burst length= 90 µsec Control burst processing time= 2.5-4 µsec Switching time = 12 µsec Propagation delay on a link = 0.2 to 1 millisecond C. Results igure 4 shows that with the proposed scheme (PSBHC) based network performs overall increasingly better than the best-effort network for increasing load. This is because a best-effort network drops a packet that has performed several hops in the network with the same probability as a new packet. Such a long distance packet may on its path have blocked other packets from being served by the switch and the resources will thus be wasted compared to PSBHC. By differentiating the packets according to their hop count level (HCL), the PSBHC scheme avoids this effect and increases the throughput. 1 Best-effort PSBHC (Proposed scheme) Loss Probability 0.1 0.01 0.001 0.2 0.4 0.6 0.8 1.0 Offered Load ig.4. Total Loss Probability in PSBHC vs. Best-effort

98 The average resource usage for dropped packets (i.e. the resource waste) will according to the above reasoning be lower in the PSBHC network, as shown in ig-5. The resource waste will decrease for an increasing load. This can be explained by knowing that the resource waste is the average number of resources a packet uses before it s dropped (i.e. average number of hops performed before being dropped). When the load increases the average number of hops decreases before the packet is dropped thus lowering the resource waste. 1.4 1.2 Best-effort PSBHC (Proposed scheme) Resource-waste 1.0 0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 0.8 1.0 Offered Load ig.5. Resource Waste in PSBHC vs. Best-effort. igure 6 shows the burst loss probability versus the offered load ρ ( α = 25). It shows that the proposed scheme offers lower burst loss probability than the conventional scheme. This is because in the proposed scheme, due to prior-information at each edge router, the traffic is not concentrated on particular link. Also, the proposed scheme with N = 2 offers lower burst loss probability than that PI with N = 0. This is because the transmission of prior-information packets enables edge routers to PI get more information and to reduce the number of discarded bursts in this process. Burst Loss Probability 1e+0 1e-1 1e-2 1e-3 1e-4 Shortest-path Proposed with N PI =0 Proposed with N PI =2 1e-5 0.2 0.4 0.6 0.8 1.0 Offered Load ig.6. Burst Loss Probability vs. the Offered Load ρ ( N PI : The number of prior-information packets). igure 7 shows the total end-to-end throughput versus the offered load intensity. The JET scheme exhibits the lowest throughput due to the path length priority effect. In contrast, the proposed scheme

99 exhibits the best performance in the end-to-end throughput Therefore, this scheme provides that a burst traversed more hops can complete its transmission more successfully. 1.0 0.8 PSBHC (Proposed) JET Throughput 0.6 0.4 0.2 0.0 0.2 0.4 0.6 0.8 1.0 Offered Load ig.7. Comparison of End-to-End Throughput Performance igure 8 shows that class 1 traffic has better performance with the proposed scheme, especially when the load of the network (in Erlang) is greater than 70. Moreover, because class 1 traffic consumes more bandwidth, it also degrades the performance of class 0 traffic. However, ig-9 shows that the total blocking probability in the network is almost the same, which indicates that proposed scheme does not degrade the blocking probability of the network. 0.1 Blocking Probability 0.01 0.001 Proposed with Class(C0) CS with Class(C0) CS with Class(C1) Proposed with Class(C1) 0.0001 20 40 60 80 100 Load( in Erlang) ig.8. Comparison of Performance of Different Traffic. 0.04 Proposed scheme CS Blocking Probability 0.03 0.02 0.01 0.00 20 40 60 80 100 Load(in Erlang) ig.9. Total Burst Loss Probability (BLP) in Proposed and CS (Time window = 0.5s, Offset time = 0.25s)

100 V. CONCLUSIONS In the proposed scheme, each edge router finds a suitable route to the destination edge router autonomously by using feedback and prior-information packets. The simulation results showed that the proposed scheme performs better in terms of loss probability and resource waste than a best-effort network. The throughput is thus increased and the network resources are used more effectively. This is especially true for high loads and few wavelengths per fiber. The proposed scheme also performs admission control. Only new packets that don t block already admissible packets are allowed into the core network. The simulation results have proved that the real time applications which are denoted by class 1 traffic have a better performance using the proposed scheme than with CS scheme. Also, the proposed scheme can avoid the path length priority effect and enhance the end-to-end throughput in multiple hop network environments. By leveraging statistical multiplexing, re-configuration is minimized and bandwidth utilization is enhanced. REERENCES [1] C. Qiao, and M. Yoo, Optical burst switching (OBS)-A new paradigm for an optical internet, J. High Speed Networks, vol.8, no.1, pp.69-84, March 1999. [2] T. Battestilli and H. Perros, An introduction to optical burst switching, IEEE Communications Magazine, Volume: 41, Issue: 8, pp. S10 - S15, Aug. 2003. [3]Amit Kumar Garg, R S Kaler, Performance Analysis of Optical Burst Switching High-Speed Network Architecture, in International Journal of Computer Science and Network Security (IJCSNS}, Vol.7, No.4, pp.292-301, April 2007. [4]Amit Kumar Garg, R S Kaler, Performance Analysis of an Integrated Scheme in Optical Burst Switched High-Speed Networks, in Chinese Optics Letters (COL)-International Optical Journal in China, Vol.6, No.4, April 2008. [5] B. C. Kim, Y. Z. Cho, J. H. Lee, Y. S. Choi, and D. Montgomery, Performance of Optical Burst Switching Techniques in Multi-Hop Networks, Proc. IEEE GLOBECOM 2002, Taipei, Taiwan, Nov. 2002. [6] Y. Xiong, and M. Vandehoute, and H. C. Cankaya, Control architecture in optical burst-switched WDM networks, IEEE J. Select. Areas In Commun, vol.18, no.10, pp.1838-1851, Oct. 2000. [7] Network Simulator, NS-2, available at http://www.isi.edu/nsnam/ns