Three PHB traffic flows of differentiated services are simulated in a mesh network topology shown in Figure 12. This is used to simulate a multiple LSPs in an enterprise network. R1 in Figure 12 is the edge router of the enterprise network whereas R2, R3 and R4 are core routers. Three LSPs are available to support differentiated services in the enterprise network. Table 5 describes the simulation parameters: type of traffic flows, the start time of flows, the guaranteed bandwidth of flows, and the delay time limit of flows. Delay time here means the maximum tolerance of the data flow that is delayed.
EF1 EF2 AF1 AF2 AF3
BE1 Router R1 Router R2
1Mbps
1Mbps 20ms
1Mbps 1Mbps 1Mbps 1Mbps 1Mbps
Router R4
Router R3 2Mbps
10ms 1Mbps
5ms
2Mbps 10ms 1Mbps
5ms
Figure 12. An enterprise network differentiated service mechanism in a multiple LSPs topology Table 5. Simulation parameters of an enterprise network differentiated service
mechanism in a multiple LSPs
Traffic flow Start time (Second) Bandwidth (Mbps) Delay time (ms)
EF1 0 1.0 15
EF2 2 0.4 20
AF1 0 1.2 None
AF2 1 0.8 None
AF3 3 0.4 None
BE1 0 0.0 None
With the network topology of Figure 12 and the simulation parameters of Table 5, simulation results are shown in Figure 13.
R1-R4-R2 LSP Transmission
0 0.2 0.4 0.6 0.8 1
0 1 2 3 4
Time（Second）
Badwidth（Mbps）
EF1
R1-R3-R2 LSP Transmissions
0 0.2 0.4 0.6 0.8 1
0 1 2 3 4
Time（Second）
Bandwidth（Mbps）
AF1 AF2
Figure 13.A Figure 13.B
R1-R2 LSP Transmissions
0 0.2 0.4 0.6 0.8 1
0 1 2 3 4
Time（Second）
Bandwidth（Mbps）
EF2 AF3 BE1
Figure 13.C
Figure 13. Simulation results of differentiated services in multiple LSPs Several observations are in order.
(1) At time t = 0, the R1-R4-R2 LSP has the minimum transmission delay of 10 ms and is the only choice for the EF1 flow to guarantee its 15 ms delay requirement.
(2) At time t = 0, the AF1 only requires 1 Mbps bandwidth. Because of the assured bandwidth of 1.2 Mbps, only the R1-R3-R2 LSP could satisfy. The path of R1-R2 is not chosen because it cannot assure the 1.2Mbps bandwidth that AF1 requires (see Table 5).
(3) At time = 0, having assigned EF1 and AF1, the network can only assign R1-R2 LSP to BE1 since that is the only LSP left. Technically, LSP R1-R3-R2 still has 1 Mbps but the assured bandwidth of AF1 is 1.2 Mbps with 0.8 Mbps left to others. This 0.8 Mbps cannot satisfy 1 Mbps of BE1.
(4) At time t = 1, the AF2 flow begins its transmission. Only the R1-R3-R2 LSP has 1 Mbps idle bandwidth. Therefore, AF2 takes this remaining bandwidth.
(5) At time = 2, all LSPs are allocated to the EF1, AF1, AF2 and BE1 flows. No idle bandwidth is available. Bandwidth preemption would occur to satisfy the transmission requirement of EF2 flow.
According to Minimum Priority algorithm, the EF2 flow selects the R1-R2 LSP as its transmission path. EF2 now takes away some bandwidth from BE1. Since the network
administrator only guarantees 0.4 Mbps to EF2 (see Table 5), EF2 must preempt 0.4 Mbps from BE1 where BE1 is now the lowest priority in all flows. Note that BE1 keeps using the remaining 0.6 Mbps bandwidth.
If the EF2 flow selects the R1-R3-R2 LSP as its transmission path, it will cause that the AF2 flow must reselect the R1-R2 LSP and preempt the 0.8 Mbps bandwidth from BE1 flow. This LSP selection of EF2 would reduce the network utilization, the 0.6 Mbps bandwidth are idle in the R1-R3-R2 LSP, and increases the transmission path overhead of AF2.
(6) At time t = 3, the AF3 traffic flow begins its transmission. By using Minimum Priority algorithm, AF3 now can only preempt BE1 since BE1 is now the lowest priority flow. The preempted bandwidth is 0.6 Mbps and is exactly the amount needed to assure AF3.
Table 6 A comparison of traffic flows’ expected transmission loss and simulation results in multiple LSPs
(unit: Mbits) 0th ~ 1st
second
1st ~ 2nd second
2nd ~ 3rd second
3rd ~ 4th second
Simulated data loss 0 0 0.598 0.594
Expected data loss 0 0 0.6 0.6
EF
Total transmission size 1 1 2 2
Simulated data loss 0 0 0 0.366
Expected data loss 0 0 0 0.4
AF
Total transmission size 1 2 2 3
Simulated data loss 0 0 0.964 1
Expected data loss 0 0 1 1
BE
Total transmission size 1 1 1 1
Table 6 lists a comparison of EF, AF and BE traffic flows’ expected transmission loss and the simulated data losses. It shows the simulated data losses of EF, AF and BE traffic flows are close to the expected transmission losses. This also demonstrates MPLS could support an enterprise network that uses differentiated service mechanism.
5. Conclusion
With advances of Internet applications, enterprises gradually adopt information technologies to enhance their business performance. The prevailing of E-commerce also rests much business burden on a responsive intranet. To survive or to compete in today’s competitive business environment, executive managers must make timely business decision. A responsive decision calls for a responsive network.
The enterprise network differentiated service mechanism may solve an insufficient bandwidth issue over an enterprise network. Differentiated service mechanisms depend on the transmission
characteristics of enterprise traffic flows to provide differentiated services among these flows. Traffic flows of enterprise can be classified into three categories: EF traffic flows, AF traffic flows and BE traffic flows. Each of these three traffic flows receives different transmission priority and network bandwidth. An EF traffic flow gets guarantied bandwidth and full transmission service quality at any time. EF traffic flows are good for transmitting urgent information. Minimum assured bandwidth would be allocated to an AF traffic flow. If there is idle bandwidth, an AF traffic flow also could get more bandwidth allocation than its assured bandwidth. AF traffic flows also provide a partial transmission service quality for an enterprise. AF traffic flows are good for delivering important operation information. The transmission priority of a BE traffic flow is the lowest. After satisfying all transmission requirements of EF and AF traffic flows, a BE traffic flow receives bandwidth if there are still available bandwidth. BE traffic flows do not receive transmission service quality. One can use BE to transmit regular messages for an enterprise.
In this research, several simulation scenarios are simulated. The simulation results demonstrate the enterprise network differentiated service mechanism could be a good choice for an enterprise to improve its enterprise network’s operation performance with the minimum investment. This would be helpful for an enterprise to enhance its competition capability and expand its business opportunities.
Acknowledge
This research is sponsor by NSC 89-2416-H-009-011
Reference
1. Andersson, L., et. al.(2000), “LDP Specification,” draft-ietf-mpls-ldp-08.txt
2. Awduche, D., et. al. (1999), “Requirements for Traffic Engineering Over MPLS,” RFC 2702 3. Callon, R., P. Doolan, N. Feldman, A. Fredette, G. Swallow and A. Viswanathan (1999), “A
Framework for Multiprotocol Label Switching,” draft-ietf-mpls-framework-05
4. Deering, S. and R. Hinden (1998), “Internet Protocol, Version 6(IPv6) Specification”, RFC 2460 5. Jamoussi, B., et. al.(1999), “Constraint-Based LSP Setup using LDP,” draft-ietf-mpls-cr-ldp-03.txt 6. J. Heinanen, F. Baker, W. Weiss and J. Wroclawski (1999), “Assured Forwarding PHB Group,”
RFC 2597
7. K. Nichols, F. Baker and D. Black (1998), “Definition of the Differentiated Services Field (DS Field) in the IPv4 and IPv6 Headers,” RFC 2474
8. K. Nichols and K. Poduri (1999), “An Expedited Forwarding PHB,” RFC 2598 9. Le Faucheur et. al.(2000), “MPLS Support of Differentiated Services”,
draft-ietf-mpls-diff-ext-05.txt.
10. Mohammad Peyravian and Ajay D. Kshemkalyani(1998), “Decentralized network connection preemption algorithms,” Computer Networks and ISDN Systems, (30), pp. 1029-1043.
11. Postel, J., Editor (1981), “Internet Protocol”, STD 5, RFC 791
12. Sally Floyd and Van Jacobson (1995), “Link-sharing and Resource Management Models for Packet Networks,” IEEE/ACM Transaction on Networking, 3(4)
13. S. Blake, D.Black, M. Carlson, E. Davies and W. Weiss (1998), “An Architecture of Differentiated Services,” RFC 2475
14. S. McCanne and Sally Floyd, “ns-Network Simulator,” URL http://www-mash.cs.berkeley.edu/ns/
15. http://www.find.org.tw/img_howmany\howmany_201.gif 16. http://www.zonaresearch.com/deliverables/white.htm