226
streaming media . It should be noted that for any given codec, as the voice payload per packet is increased while the net bandwidth use is reduced. Fewer packets are needed to transmit the same amount of data that reduce the net overhead [7]. The calculations are computed as follows: Voice packet size = layer 2 header + (IP + UDP + RTP) Header + voice payload. Voice packets per second = codec bit rate / voice payload size Bandwidth = voice packet size x voice packets per second. In this experiment, compression of IP, UDP and RTP headers for G.711 (64) codec are not taken into consideration. Thus the UDP packet size in bytes = UDP header + RTP header + voice payload = 8 + 12 + 160 = 180bytes. For the Largest Packet Size, Ethernet MTU size packets are considered. Ethernet MTU = 1500bytes. Of these 1500 bytes, the IP headers use 20bytes, leaving 1480 bytes for the UDP packet (including header and payload). Thus UDP packets of 1480bytes are considered. And for creating worst scenario, another UDP packet of 830 bytes and Drop Tail queue type have also been considered. III. 3. SIMULATION AND PERFORMANCE ANALYSIS We have used NS2 to create the topology as shown in figure 1 for both traditional and MPLS networks. The simulation has been developed to emphasize the impact of Traffic Engineering over the traditional network. Nodes 0 and 1 are used as source and nodes 6 and 7 are used as destination.
The information table 1 of node LSR2 act as Label Edge Router (LER) shown below, where egress router characterize FEC, label and LSPID. Labels of this table are distributed based on the control mode that is chosen to be executed at node LSR2. During packet transmission, label swapping is done by each intermediate node. Routing information is stored in LIB, PFT and ERB tables using mapping message from egress router to ingress router(LSR2 to LSR7) and vice versa.
Figure 2: Symmetric Network without MPLS In MPLS network, node 2 to node 5 is defined as LSR nodes. We have considered LSR2 as ingress and LSR5 as egress where the path through node 2_3_5 considered as shortest path. Due to the congestion at node 2, traffic engineering is applied for MPLS network. Traffic follows the alternative path (via node 2_3_5).
Figure 3: Symmetric Network with MPLS Figure 1: Simulation Topology In traditional network path (via node 2_4_5) is not utilized, while path (via node 2_3_5)is over utilized. Packets are lost due to congestion at node 2 and we have observed same situation when no traffic engineering is applied for MPLS network. A. ANALYTICAL REPORT The figure 4 below shows the throughput of received packets (MB) for each flow of traffic from source nodes before applying traffic engineering where received level is approximately 0.9 MB for node 6 and 0.6 MB for node 7.
227
TABLE I.
PFT dump___ [node: 2] --------------------------------------------FEC PHB LIBptr AltanativePath 7 -1 0 -1 ERB dump___ [node: 2] --------------------------------------------FEC LSPid LIBptr 5 3600 0 5 3700 1 LIB dump___ [node: 2] --------------------------------------------# iIface iLabel oIface oLabel LIBptr 0: -1 -1 4 1 -1 1: -1 -1 4 2 -1
After applying the Traffic Engineering in MPLS, network congestion as well as packet received level is improved. Traffic follows the path (via nodes 2_4_6) as an explicit route. Throughput of MPLS network shown in figure 5 where received level is approximately 1.4 MB up to 3 sec then the curve slowly goes down at 4.5 sec it is almost 1 MB for node 6 and almost 1 MB for node 7.
TABLE II. COMPARISONS OF PACKETS (IN BIT) BEHAVIOR BETWEEN TRADITIONAL AND MPLS NETWORK
Packet type Total CBR Packet send Total CBR Packet receive Total CBR Packet drop
Figures (6 and 7) below show (graphically) the number of packet drop behavior for both Traditional and MPLS network with respect to time. Figures (6 and 7) indicate the rapid fluctuation of packets. The average number of packets drop are approximately 6 and 2 for traditional and MPLS network respectively.
Figure 7:Packet drop Behaviors of MPLS Network Figure 5: Throughput of a MPLS network.
Table given below shows the total number of packets transmitted from sources, received by
228
destination nodes and dropped due to congestion for both traditional and MPLS network. Packet loss is much less in MPLS network than that of traditional network. IV. CONCLUSION This paper has been prepared based on the traffic flow over both conventional and MPLS network, where network topology and other simulation parameters are chosen as common to establish the better performance of MPLS network over traditional network. Based on the comparison of signaling protocols, it can be found that RSVP has drawback in its scalability that is one reason for choosing CR-LDP for MPLS protocol. The results are obtained after some experimentation and calculation with network scale (number of nodes, link capacity and delay) and traffic arrangements (sources and packet sizes, and CBR packet arrival rates). As expected, packet transmissions (in terms of both delay and loss) are improved in MPLS network. Although the chosen parameters can
be disputed to be artificially extreme, the traffic engineering mechanism improves the performance of general delay and loss. REFERENCES
[1] [2] [3] [4] [5] [6] [7] Analysis of MPLS traffic engineering - Circuits and Systems, 2000, www.ieeexplore.ieee.org/iel5/7554/20602/00952816.pdf Independent study report Study of Traffic Engineering Algorithms, www.ieee.unlv.edu/~venkim/opnet/IndependentStudy.pdf Signaling Protocols in DiffServ-aware MPLS Networks: Design and Implementation of RSVP-TE Network Simulator, www.ieeexplore.ieee.org/iel5/10511/33286/01577748.pdf L. Andersson et al. LDP Specification, IETF RFC, 3036, January 2001 B. Jamoussi et al. Constraint-Based LSP Setup using LDP, IETF RFC,3212, January 2002 D. Awduche et al. RSVP-TE: Extensions to RSVP for LSP Tunnels, IETF RFC 3209, December 2001 Voice over IP - Per Call Bandwidth Consumption. http://www.cisco.com/warp/public/788/pkt-voicegeneral/bwidth_consume.html
229