European Journal of Scientific Research ISSN 1450-216X Vol.25 No.1 (2009), pp.54-67 © EuroJournals Publishing, Inc. 2009 http://www.eurojournals.com/ejsr.htm
Traffic Engineering: Simulation Model and Real Network Environment Over Single and Multiple Links
Mohd Nazri Ismail Faculty of MIIT, University of Kuala Lumpur, Malaysia E-mail: mnazrii@miit.unikl.edu.my Abdullah Mohd Zin Faculty of FTSM, University of Kebangsaan Malaysia, Malaysia E-mail: amz@ftsm.ukm.my Abstract We present a novel approach for the measurement and estimation of single and multiple links in heterogeneous network environment. We propose an enhanced equation to evaluate the performance of network link via Little Law and Queuing theories (M/M/1 and M/M/m) to improve the evaluation algorithm. To get accuracy results on the performance of traffic engineering simulation model, we verify and validate data from real network environment. We use network management tool to capture those data and network analyzer device to generate traffic into single and multiple links. As a result, this traffic engineering simulation model can provide a good approximation of the real traffic observed in the real network environment. Keywords: Traffic engineering, simulation model, multiple links, single link, network management
1.0. Introduction
Traffic engineering of the network is performed to optimize the network performance of a telecommunications network by dynamically analyzing, predicting and regulating the behaviour of data transmitted over the network such as reducing traffic congestion in the network. Traffic engineering is also known as tele-traffic engineering and traffic management. Considerable research has been conducted to model and quantify the performance of heterogeneous services and technologies (e.g. [1], [2]. Several flow-level network traffic models have been proposed to describe/stimulate [3], [4]. These models have been used to study fairness, response times, queue lengths and loss probabilities under different assumptions and using a variety of mathematical techniques. In contrast to other works in the literature (e.g., ([5], [6]), we developed traffic engineering simulation model to measure the performance of heterogeneous environment over single and multiple links. The significant of this study was to develop a traffic engineering simulation model to measure the performance of single and multiple links in heterogeneous network environment using Queuing theory. The objective of traffic engineering in heterogeneous environment is to avoid congestion in the network and make better use of available resources. This traffic engineering simulation model is designed to: i) define appropriate traffic and congestion controls to safeguard performance in case of failure or overload the links; ii) assist network administrator to prepare, propose, plan and design
Traffic Engineering: Simulation Model and Real Network Environment Over Single and Multiple Links
55
network topology more effective and systematic; iii) conduct ‘What-If’ analysis for evaluating single and multiple links in heterogeneous environment performance; iv) integration of networks with different technologies and providing different services; v) emerging of new multimedia such as IP telephony; and vi) due to major changes in access network technologies (such as cable modem, xDSL and digital line) for incorporating new services. Users are now demanding and expecting more services. Many factors may contribute to the congestion of network interface, such as a heavy load in the network that usually generates higher traffic [7]. Thus, this research is critical to be conducted in order to predict and measure of traffic engineering in heterogeneous environment.
2.0. Related Works
In the 21 century, a network infrastructure is based on multi-service implementation over convergence of network medium such as ISP, PSTN and GSM [8], [9]. Multi-traffic in the network infrastructure has become more complex to observe and analyze [10]. Local Area Networks (LANs) are exposing the impact as more users and powerful applications threaten to overwhelm network capacity [11]. The main factors of network congestion are related to network design and bandwidth capacity [12]. Mechanisms are needed that can handle traffic load dynamically distribute traffic to benefit from available resources [13]. Today, traffic is usually routed on the single link through a network. Many tools have been developed, and only a few tools have successfully achieved a close estimation of network bandwidths. We investigate how optimization techniques can be applied to heterogeneous environment in order to better utilise network link and to avoid congestion by balancing load over multiple links. We have setup a real network environment to analyze and measure of network traffic utilization at University of Kuala Lumpur in Malaysia. This study posits several research questions: i) what is the performance level of the network utilization over single and multiple links; and ii) Is the traffic engineering simulation model for evaluating and measuring single link and multiple links in heterogeneous environment performance effective?
3.0. Methodologies
The techniques of traffic engineering can be applied to networks of all kinds, including the PSTN, LANs, WANs and cellular telephone. Whatever modelling paradigm or solution techniques in heterogeneous environment model development are being used, the performance measures extracted from a simulation model must be a good representation of the real network environment. Figure 1 shows the overall framework of the traffic engineering simulation model. There are three performance techniques to validate the traffic engineering simulation model: i) tracing; ii) parameter variability; and iii) predictive validation. In addition, there are two techniques to judging how good a model is with respect to the real network: i) model verification; and ii) model validation. Comparison with a real network is the most reliable and preferred method to validate a simulation model (refer to Figure 2). Figure 3 shows the mathematical model validation process has conducted to ensure the accuracy of algorithms. Figure 4 and Figure 5 show the network design are based on single and multiple links. While, Table 1 shows the legend for single and multiple links network design.
56
Mohd Nazri Ismail and Abdullah Mohd Zin
Figure 1: Simulation Model Development Methodology
Figure 2: Simulation Model Verification and Validation Methodology
Table 1:
Legend PC S CS R I SV
Legend for Single and Multiple Links
Remarks Computer (PC1, PC2, PC3….PCn) Switch Core Switch Router (R, R1 and R2) Internet Server
Traffic Engineering: Simulation Model and Real Network Environment Over Single and Multiple Links
Figure 3: Mathematical Model Validation Process
57
Figure 4: Network Design for Single Link
Figure 5: Network Design for Multiple Links
58
Mohd Nazri Ismail and Abdullah Mohd Zin
4.0. Propose Traffic Engineering Simulation Model Development
Many different types of modeling and simulation applications are used in various disciplines such as acquisition, analysis, education, entertainment, research and training [14]. In the Figure 4.1, theoretical model is based on a random distribution of service duration. Simulation model is divided as follows: i) to study physical of real heterogeneous network environment; ii) transform physical of real heterogeneous network environment into logical model; and iii) develop and implement the traffic engineering simulation model in heterogeneous environment. Physical Model of Single and Multiple Links in Real Network Environment Figure 6 shows the network heterogeneous environment in real world. Then we need to transform from heterogeneous environment in real world into logical model. The logical model is the phase where mathematical techniques are used to stimulate traffic engineering in heterogeneous environment.
Figure 6: Real Heterogeneous Network Environment
Logical Model of Traffic Engineering in Heterogeneous Environment The logical model in original Queuing theory (M/M/1 and M/M/m) will be derived to meet requirement for complex situation in heterogeneous environment (see Figure 7). From the original Queuing theories will use to develop logical model of traffic engineering in heterogeneous environment for network traffic utilization (see Figure 8). Parameters like bandwidth capacity, size of packet services and number of clients are used to ‘characterize’ the application traffic.
Figure 7: Logical Model for Original Queuing Theory
Traffic Engineering: Simulation Model and Real Network Environment Over Single and Multiple Links
59
Figure 8: Logical model of Single and Multiple Links (Traffic Engineering) in Heterogeneous Environment
Development of Traffic Engineering in Heterogeneous Environment Model This section describes a simple analytical queuing and little law theories that capture the performance characteristics of network utilization and traffic operations. A link refers to a single connection between routers and hosts. The link bandwidth is the rate at which bits can be inserted into the medium [15]. Table 2 shows the parameters that have been used in the model development. In open queuing network, the traffic of the heterogeneous network environment is determined by the input rate in the system. Table 3 summarizes all the parameters used in the model.
Table 2: Notations for Original Queuing and Little Theories
Meaning Average number of clients in the system Average time a client spends in the system (second) Clients arrival rates Service rate in second Mean service times Multiple links Traffic intensity
Model Parameters N T
λ
ρ
μ 1/μ m (ρ12..m)
60
Table 3: Notations for Model Development
Mohd Nazri Ismail and Abdullah Mohd Zin
Model Parameters N T P (P1,P2,P3,...Pm) P1 µclient
μ server
Nclient + server U client + server C(CLAN, CWAN) (C1,C2...Cm) Uhetergenes Heter client + server µTotal Tot_client Tjum µJum_S µSer_S
Meaning Average number of traffics on the network Average time of clients arrive on the network (second) Various of services Client uses single service Size of packet service request by client (traffic) Traffic response from server to clients Number of clients in second over single service Network traffic utilization usage based on number of clients in second over single service Size of Bandwidth on LAN and WAN interface ports Network traffic utilization usage for heterogeneous environment Number of clients and traffics over heterogeneous environment Total size of packet services request by clients (traffic) Number of clients Total number of clients access on the network (1 sec) Total size of new packet services New packet services
The original Queuing theory is defined as an average number of clients in the system (variable name is ‘N’) in equation 1. Equation 4 is defined as traffic intensity use by clients in the system. Equation 1, 2, 3 and 4 are derived based on logical model that has been designed to meet requirements for heterogeneous network environment. Logical model is derived and formulated in a single service (homogeneous concept) as in equations 5, 6, 7, 8 and 9. Then, the logical model is derived to the single link heterogeneous environment in equations 10, 11, 12, 13, 14 and 15. Then, single link logical model is derived to multiple links for heterogeneous environment in equations 16, 17, 18, 19 and 20. Rigorous formulation is based on Figure 4.2: The system is characterized by {(An, Dn), n ≥ 1}, where An – time of the nth arrival Dn – departure-time of the nth arrival 0 ≤ An ≤ An + 1 An ≤ Dn ≤ ∞ Define: A(t) = number of arrivals until t; D(t) = number of departures until t; i.e., number of units in the N(t) = number of units such that; An ≤ t ≤ Dn system; Tn = Dn – An, sojourn time of the nth unit in the system. Theorem. Assume that 1 1 A ( t ) = lim λ = lim D ( t ), 0 〈 λ 〈 ∞ t→ ∞ t t→ ∞ t Then
1 N lim ∑ Tn = T N →∞ N n =1 ⇔ 1 T ∫ N (t) = N T →∞ T 0 lim
In which case
N = λ * T
T = 1 μ
(1) (2)
Traffic Engineering: Simulation Model and Real Network Environment Over Single and Multiple Links
N = λ 1 μ = ρ
61
(3)
ρ=
λ < 1; λ < μ μ
(4)
Single Link: Client uses single service for accessing network server N = µclient * (Tot_client * T) Nclient + server = P1 * (Tot_client * T) Nclient + server = (Request+Response)*(Tot_client*T) Nclient + server = (µclient + µserver)*(Tot_client*T) Uclient + server = Nclient + server / C Single Link:
(5) (6) (7) (8) (9)
Client uses various services for accessing network server in Heterogeneous Environment (10) TJum = Tot_client * T Heterclient + server = (P1+P2+P3+….+ Pm)*TJum (11) Heterclient + server = (µclient1+µclient2+µclient3 +..+ µclient_m) * TJum (12) Heterclient + server = (µTotal) + (µserver) * TJum (13) Where P1+P2+P3+…+Pm = Uclient1+Uclient2+Uclient3+…+Uclient_m = µTotal (14) Uhetergenes = Heter client + server / C (15)
Multiple Links: Client uses various services for accessing network server in Heterogeneous Environment
ρ = λ m μ
(16) (17)
=
=
λ
m ( Router 1 + Router 2 + ... + Router m )
(18) (19)
λ
m (C1 + C 2 + .. + Cm )
T*[(µTotal +µserver)+(µJum_S +µSer_S)]
ρ12..m =
m*(C1 + C2 + …..+ Cm)
(20)
5.0. Accuracy of Simulation Model with Real Network Experimental
In this section, we verify the little law and queuing theories for traffic engineering simulation model in heterogeneous environment through experiments. Real experiment is based on real network and need to consider as follows: i) network bandwidth is limited and is not enough for all application and users at the same time; ii) delay due to the network overloads; and iii) packet losses. Real Network Setup We used a network management application to capture traffic between single and multiple links in real network environment. Figure 9 and Figure 10 show the experimental setup of real network used in our
62
Mohd Nazri Ismail and Abdullah Mohd Zin
tests. Fluke Optiview (network analyzer) device can be configured to insert size of packet services and number of clients to generate traffic into single and multiple links (see Figure 11). By using varying number of clients and size of packet services, we are able to simulate network utilization and traffic between single and multiple links. Figure 10 shows the server is link to Ethernet 100 Mbps and the second link is WLAN 54 Mbps, see Table 4 and Figure 12. We will demonstrate that multiple links can reduce network congestion and overhead. This technique is very useful in traffic engineering and we will simulate via simulation model.
Figure 9: Real Network Environment Setup for Single Link to Internet
Figure 10: Real Network Environment Setup for Multiple Links to Internet
Traffic Engineering: Simulation Model and Real Network Environment Over Single and Multiple Links
Figure 11: Fluke Optiview Engine Setting
63
Figure 12: Network Management Configuration
Table 4:
Server Configuration and Specification
Server, Link 1 (Ethernet LAN) Broadcom NetLink Gigabit Ethernet 00-17-08-41-B1-80 10.5.1.11 10.5.1.254 100 Mbps Server, Link 2 (WLAN) Intel® PRO/Wireless 3945ABG Network 00-18-DE-7B-AF-24 10.40.0.247 10.40.1.254 54 Mbps
Specification Network Card MAC IP Gateway Bandwidth
Traffic Engineering - Real Network Experiment and Simulation Model We have setup a real network environment of network utilization measurement that generates measurement data to analyze network performance at the main campus. We pump traffics 3.2 MB in parallel into LAN 100 Mbps (first link) and WLAN 54 Mbps (second link) (see Figure 13, 14 and 15). We run network management application to measure traffic and its network utilization performance between single and multiple links (see Figure 13, 14 and 15).
64
Mohd Nazri Ismail and Abdullah Mohd Zin
Figure 13: Utilization: Multiple Links via LAN and WLAN Connectivity
Figure 14: Utilization: Single Link via WLAN Connectivity
Figure 15: Utilization: Single Link via LAN Connectivity
Traffic Engineering: Simulation Model and Real Network Environment Over Single and Multiple Links
65
Two sets of experiments were conducted with different scenarios (see Table 5, 6, 7 and 8). We used the same input variables (3.2 MB) that have been used in real network environment to estimate our result that must be closely resemble to traffic engineering simulation model between single and multiple links (refer to Figure 16). Table 5 and Table 6 show the comparison and relative error rates between simulation model and real network environment for single link (refer to Figure 16). Then, Table 7 and Table 8 show the comparison and relative error rates using traffic engineering techniques between simulation model and real network environment for multiple links (refer to Figure 16).
Figure 16: Traffic Engineering Simulation Model for Single and Multiple Links over Wide Area Networks
Table 5:
Experiment 1: Utilization of Single Network Link
Input: Total Size of packet services = 3.2 MB LAN A: 1.6 MB; LAN B: 1.6 MB Utilization(0-1%) Utilization(0-1%) (Minimum Range) (Maximum Range) 0.05 0.30 0.03 0.21
Single Link LAN (100Mbps) WLAN (54 Mbps)
Simulation Model (0-1%) 0.256 0.474
Table 6:
Experiment 2: Utilization of Single Network Link
Input: Total Size of packet services = 3.2 MB LAN A: 1.6 MB; LAN B: 1.6 MB Utilization (0-1%) Utilization(0-1%) (Minimum Range) (Maximum Range) 0.02 0.35 0.08 0.5
Single Link LAN(100Mbps) WLAN(54 Mbps)
Simulation Model (0-1%) 0.256 0.474
We conclude that base on our findings, the traffic engineering simulation model for multiple links are able to predict and estimate network utilization usage for real network environment with minimum relative error rates (see Table 7 and Table 8). Table 5 and Table 6 show single network link can generate higher network utilization rate compare to multiple links in heterogeneous environment.
66
Table 7:
Mohd Nazri Ismail and Abdullah Mohd Zin
Experiment 1: Utilization of Multiple Network Links
Input: Total Size of packet services = 3.2 MB LAN A: 1.6 MB; LAN B: 1.6 MB Utilization(0-1%) Utilization (0-1%) First and Second Links (154 Mbps) (Minimum Range) (Maximum Range) LAN 0.02 0.12 WLAN 0.01 0.05 LAN + WLAN 0.03 0.17
Simulation Model (0-1%) 0.0831
Table 8:
Experiment 2: Utilization of Multiple Network Links
Total Size of packet services = 3.2 MB LAN A: 1.6 MB; LAN B: 1.6 MB Utilization(0-1%) Utilization(0-1%) (Minimum Range) (Maximum Range) 0.008 0.14 0.02 0.07 0.028 0.21
First and Second Links (154 Mbps) LAN WLAN LAN + WLAN
Simulation Model(0-1%) 0.0831
6.0. Conclusion
In this article, we have shown how an analytical queuing network model can be used to understand the behaviors of heterogeneous environment over single and multiple links experiments. Our traffic engineering simulation model, has demonstrated that it can measure accurately the performance of heterogeneous services and technologies over multiple links. We believe the traffic engineering simulation modeling framework described in this study can be used to study other variations, tunings, and similar new ideas for various services and technologies. In network management, by monitoring and analyzing network utilization rate we can monitor the performance of the network, thus to study whether network is normal, optimal or overloaded. Therefore, if the first network link is overloaded then multi-traffic can route to the second network links.
Traffic Engineering: Simulation Model and Real Network Environment Over Single and Multiple Links
67
References
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] Kawasaki, S., M. Niwa, T. Kamina, D.A. Wu and H. Murakami. 2006. A Study on Formulation of the Ubiquitous Cloud Model. 7th International Conference on Mobile Data Management (MDM’06), May 10-12, pp:148 – 148. Tsalgatidou, A., G. Athanasopoulos, M. Pantazoglou, C. Pautaos and T. Heinis, 2006. Developing scientific workflows from heterogeneous services. ACM Sigmod, 35: 22 – 28. Guojun, J. and B.L. Tierney, 2003. System capability effects on algorithms for network bandwidth measurement. Proceedings of the 3rd ACM SIGCOMM conference on Internet measurement (IMC '03), October 27 - 29 pp: 27-38. Fred, S.B., T. Bonald, A. Proatiere, G. Regnie and J.W. Roberts, 2001. Statistical bandwidth sharing: A study of congestion at flow level. Comput. Commun. Rev., 31: 111-122. Barakat, C., P. Thiran, G Iannaccone, C. Diot and P. Owezarski, 2002. A flow-based model for Internet backbone traffic. Proceedings of the 2nd ACM SIGCOMM Workshop on Internet Measurement, Marseille, France, 6-8 Nov., ACM, New York, USA., pp: 35 – 47. Bu, T. and D. Towsley, 2001. Fixed Point Approximation for TCP behavior in an AQM Network, ACM Sigmetrics Performance Evaluat. Rev., 29: 216-225. Kontio, J. and R. Conradi, 2002. Evaluating the Performance of a Web Site via Queuing Theory. Software Quality - ECSQ 2002: 7th International Conference, Helsinki, Finland, Springer-Verlag Berlin Heideberg, pp. 63-72. Lee, K.H., K.O. Lee, K.C. Park, J.O. Lee and Y.H Bang, 2003. Architecture to be deployed on strategies of next-generation networks. Communications ICC '03. IEEE Int. Conf., 2: 819 – 822. Li, J. and Z. Sun, 2004. Internet/Web technology in higher education in China. Proceeding of the higher education in China on Advanced Learning Technologies, August 30- September 01, 2004. IEEE Computer Society Washington, DC., USA., pp: 993 – 997. Xianxin, J., F. Yang and H. Zou, 2003. A novel architecture to customer service management for the NGN. International Conference, ICCT 2003, 9-11 April 2003, 1: 123- 126. Guo, H., Z. Jingli, O. Yangkai, Y. Shengsheng, 2004. A Design and Evaluation of Ethernet Links Bundling Systems. Proceedings of the 18th International Conference on Advanced Information Networking and Applications, AINA '04, IEEE Computer Society Vol. 2, pp: 313. Curtis, J. and T. McGregor, 2001. Review of bandwidth estimation techniques. Proceeding of the New Zealand Computer Science Research Students' Conference, Vol. 8, April 2001, University of Canterbury, New Zealand, pp: 172-174. Anders, G. and A. Henrik, 2005. Ambient Networks Management Challenges and Approaches. Vol. 3284/2004, Springer Berlin / Heidelberg, pp: 196-216. Gerhan, D.R. and S.M. Mutula, 2005. Bandwidth bottlenecks at the University of Botswana: Complications for library, campus and national development. Library Hi Tech, 23: 102-117. Sommers, J. and P. Barford, 2004. Self-Configuring Network Traffic Generation, Proceedings of the 4th ACM SIGCOMM Conference on Internet Measurement, October 25-27, ACM, New York, USA., pp: 68-81.