Proceeding of the 3rd International Conference on Informatics and Technology,

QoS and Throughput Optimization in Next Generation IP Networks Using MPLS Traffic Engineering Techniques

Muhammad Tanvir

1, Abas Md Said

2

1 Computer and Information Sciences Department,

Universiti Teknologi PETRONAS, 31750 Tronoh Perak, Malaysia. Email: [email protected] 2 Computer and Information Sciences Department,

Universiti Teknologi PETRONAS, 31750 Tronoh Perak, Malaysia. Email: [email protected]

ABSTRACT

More telecommunication services are using Internet as single network for voice, video and data transmission. Quality of Service guarantee is necessary for such transmission over the internet backbone. Internet is not a simple best-effort network only for web traffic anymore. It is generally understood that present internet best-effort infrastructure is not sufficient to provide QoS-guaranteed services. Many new protocols have been developed and optimized to realize the goal of converged services on internet. Differentiated Services (DiffServ) and Multiprotocol Label Switching (MPLS) traffic engineering are strategies that have been devised to support this transition. MPLS DiffServ traffic engineering has been utilized to support required QoS. However, focus of many QoS models has been the fair allocation of bandwidth for each class of DiffServ. Whereas, only fair allocation of bandwidth to each DiffServ class is not a comprehensive solution. Addressing these open research issues, we are carrying out study to device a new model, which would reduce delay, jitter and loss to increase the network throughput. Achieving these objectives will be a leap to realize the dream of making internet a single converged multiservice network. Keywords: DiffServ, MPLS, Next Generation Network, Traffic Engineering 1.0 INTRODUCTION

The Internet these days is undergoing a rapid change from simple point-to-point best-effort communication toward a

multiservice network that supports a number of multimedia applications and services with prospectively greater

bandwidth need.

The development of increasingly high bandwidth links has created possibilities for Internet Service Providers to go for a

tactic of bandwidth over provisioning in the networks. On the other hand, this technique presently applies only to the

core network and the need from rapidly increasing end-user IP traffic over the Internet as a whole still cannot be met [5].

The calculation results published in [4] show that bottlenecks of the Internet backbone are not only seen at inter-domain

links between autonomous systems, but also within separate domains. With this knowledge, it still is necessary for

Internet service providers to do effective resource optimization both intra and inter-domain levels, so as to remove these

bottlenecks. Internet traffic engineering is the strategy of doing this job.

Two main problems that have shortly got notice in traffic engineering techniques are QoS and robustness [5]. First, a lot

of the recent multimedia applications and services not only have bandwidth requirements, but also require other QoS

assurances, like end-to-end delay, jitter or packet loss likelihood. These QoS requirements put new challenges on

Internet service providers. Therefore the end-to-end QoS requirements should be fulfilled by traffic engineering

mechanisms. Second, due the fact that network node and link failure are still common things on the Internet, traffic

engineering mechanisms considers how to reduce the impact of failures on network performance and resource

utilization.

To the best of our knowledge, most of the work done so far in the field of IP traffic engineering is on the optimization of network utilization under different network conditions like heavy load, etc. Very little work has been done in addressing the issue of end-to-end QoS assurance. Even researchers who have addressed the end-to-end QoS have presented enhancement and optimization techniques focusing on division of bandwidth resources among multiple competing DiffServ traffic classes. As more and more delay-, jitter- and loss-sensitive Internet applications are developed to meet the sudden shift of voice, video and mission critical data services from legacy Public Switched Telephone Network (PSTN), Asynchronous Transfer Mode (ATM), and Frame Relay Networks to the world of IP (Internet). The development of new, and optimization of existing QoS assurances models, is the order of the day. MPLS has emerged as a single protocol which provides platform to comply QoS assurances. MPLS traffic engineering and DiffServ, when combined together, enable researchers to achieve bandwidth assurance models. But these models do not completely provide such QoS assurances. Therefore, it is the right time to devise QoS assurance models which also satisfy the needs of delay-, jitter- and loss-sensitive Internet services along with services with high bandwidth requirements. This can be achieved using MPLS DiffServ Traffic Engineering. Some of the work already done addressing QoS is presented in [1, 2 and 3].

©Informatics '09, UM 2009 RDT2 - 57

Proceeding of the 3rd International Conference on Informatics and Technology, 2009

Moreover, as internet is growing at a great pace, the need to optimize the Internet traffic will remain a research catching

issue in the predictable near future. This issue has not and will not be solved merely by bandwidth over provisioning.

Congestion avoidance and throughput optimization models need also be worked out.

2.0 RESEARCH PROBLEMS Quality of Service (QoS) assurance for IP traffic has been research focus for the last many years. Researchers have worked on many different aspects of QoS. Although substantial successes have been achieved in many aspects, there are still more to be done due to the everyday changing nature of IP traffic. Internet has to handle more delay-sensitive IP traffic, real-time video transmission in the form of video conferencing and web streaming are increasing parts of IP traffic; just like voice, which has been shifted from traditional Public Switched Telephone Network (PSTN) to IP networks by placing voice gateways and soft switches. Mission critical data traffic needs to be transmitted in real time between different hosts and systems across the IP network. Latency or delay should be minimal during transmission from source to destination. Secondly due to the multimedia nature of IP traffic, jitter is another QoS parameter which needs to be addressed. In supporting QoS for voice and video IP traffic, variations in delay between packets while transmission from source to destination need to be kept at the minimum or ideally constant. In practice, although it cannot be kept constant, it should be minimized to ensure end-users satisfaction. Packet loss is also one of the important QoS parameters, which, if not addressed properly, can adversely affect the overall QoS. Keeping packet loss at minimum is important to avoid retransmissions and thus degrading QoS. 3.0 RELATED WORK In this section, we present recent research work done on different aspects of Intra-Domain IP/ MPLS traffic engineering.

3.1 Intra-Domain MPLS-Based Traffic Engineering

The Internet Engineering Task Force has come up with MPLS, which is a standardized forwarding scheme. In MPLS,

traffic is sent along explicit Label Switched Paths (LSPs). An LSP is the path between an ingress label switching router

(LSR) and an egress LSR. At the boundary of an MPLS domain, LSRs classify IP packets into forwarding equivalence

classes (FECs) and add labels for packet forwarding within the MPLS domain. The Label Distribution Protocol [6] and

RSVP are used to distribute label bindings during the setup of an LSP. MPLS is a powerful technology for Internet traffic

engineering as it allows traffic to be forwarded onto random explicit route, which may not essentially follow the shortest

path computed by traditional IP Routing Protocols. Normally, each flow is aggregated by MPLS-based traffic engineering

into traffic trunks identified by FECs, which are then carried on LSPs between ingress and egress routers. In this case

the traditional shortest-path-based routing protocols (e.g., Open Shortest Path First, OSPF) are superseded with MPLS

explicit routing tunnels.

3.2 Offline Traffic Engineering

A generalized MPLS routing optimization can be devised as a multi-priority flow problem [7], and so that it can be solved using linear programming to produce an optimal solution for routing mechanisms that allow arbitrary traffic splitting. However, this method is frequently considered as impractical, particularly in a large-scale network, as the number of LSPs needed is would be enormous due to random traffic division. To get a more scalable traffic engineering solution, traffic division has to be limited in scope. An early MPLS-based traffic engineering method used simple constraint-based routing (CBR) [8] without coordination between individual traffic trunks [9]. One of such CBR algorithm is as follows. As an LSP being set for a particular traffic trunk, the links which do not fulfill the constraints are removed from the topology database. Then, shortest path routing (SPR) is then performed on the remaining network topology, and the shortest path is assigned to the LSP. These steps are repeated until all trunks are assigned to LSPs. This routing algorithm is called Constrained Shortest Path First (CSPF) routing. Some optimizations have also been proposed to add new capabilities like, Widest Shortest Path (WSP) and Shortest Widest Path (SWP) [10, 11]. These two algorithms somewhat optimize the bandwidth availability at congested links. Application of WSP/SWP, increases the chances of network traffic to find a feasible path and also congested links are avoided by sparing bandwidth for upcoming needs, rest of traffic is benefited from this routing technique. There are many MPLS-based traffic engineering techniques have been presented to reduce the maximum link utilization. In [12] traffic engineering has been evaluated by using single and multiple paths.

With the development of differentiated services (DiffServ), DiffServ-based traffic engineering has become a research area for supporting QoS differentiation. DiffServ-MPLS-based traffic engineering has now been supported by many Vendor routers, CSPF being the basic routing algorithm. More complex DiffServ aware/equivalent MPLS-based traffic engineering schemes have also been proposed [13–15]. In [14] a general outline for intra-domain QoS assurance by MPLS-based traffic engineering in DiffServ networks.



3.3 Online Traffic Engineering

Online MPLS-based traffic engineering can be categorized into two distinct categories: dynamically adjusting the traffic division ratio among pre-constructed static LSPs [29, 30], and computing dynamic LSPs in real time for each new traffic trunk demand. MATE [16] is a typical example of the first category, and its basic procedure is to adaptively send incoming traffic onto multiple pre-constructed LSPs as per network core traffic statistics. In this traffic engineering model, routing optimization is not directly addressed, and optimization is gained by online sending adaptation. CSPF, WSP, and SWP algorithms are the basic routing methods that can be applied to online MPLS-based TE strategies. Dynamic Online Routing Algorithm (DORA) [17] through the online traffic engineering method is performed in two stages; stage one is performed whenever there is any topology changes in the network. And second stage path is performed whenever there is a request for LSP setup. In the first stage a Path Potential Value (PPV) is assigned to each link for each source to destination node pairs. PPV is the parameter which shows the number of times a link can be used for future path setup request. The higher PPV value indicates the higher are the chances for a link to be utilized for multiple future paths, so to avoid this link to be allocated for future path requests. Finally, a conventional Dijkstra’s shortest path algorithm is applied based on the set of defined link weights. Simple Minimum Interference Routing Algorithm (SMIRA) [21] resolves the issue of LSP interference, setting up LSPs using CSPF without taking the location of ingress/egress nodes pairs into consideration. It is expected that congestions may occur in a few links that are used by multiple LSPs. So SMIRA, and before that MIRA, solved this problem by avoiding critical links while setting up of LSP, and spare these critical links for future traffic that will use critical link as the only path for Ingress/Egress pairs. This algorithm first distinguishes critical links for individual ingress/egress pairs by calculating the maxflow value. After that, weight is calculated for each ingress/egress pair, as a rising function criticality of the link. Finally, CSPF algorithms are applied to the network topology containing only feasible links that can support the bandwidth demand of the incoming traffic. Online MPLS-based traffic engineering has also been studied in DiffServ environments for QoS support, a typical example being traffic engineering automated manager (TEAM). The Traffic Engineering Tool (TET) in the TEAM framework is responsible for LSP preemption and construction. First, for each incoming demand, three types of cost are considered in the cost function: bandwidth, switching, and signaling. The objective of LSP manipulation is to minimize the overall cost throughout the process, which can be achieved by a Markov-process-based decision. There are two distinct LSP operations in TEAM: LSP preemption and LSP routing. LSP preemption allows existing LSPs to be preempted by newly constructed LSPs with higher priority. To do this, each LSP is assigned a priority attribute, which is taken into account when there is competition for resources (i.e., interference). Thus, even if an LSP has already been assigned a path, it will be rerouted if it has a lower priority attribute than a new LSP that is competing for the shared network resources. In order to avoid frequent LSP switching and thus traffic instability, the proposed preemption policy includes the following three guidelines: preempt the LSP with the lowest priority attribute, preempt the fewest number of LSPs, and preempt the least amount of bandwidth while satisfying the traffic demand requirement. For LSP routing, the Stochastic Performance Comparison Routing Algorithm (SPeCRA) is adopted in TEAM. SPeCRA behaves like a homogeneous Markov chain where the optimal routing scheme is a state of the chain that is visited at the steady state. Specifically, it attempts to select adaptively the best routing algorithm from a set of candidate schemes, each of which might be suitable for a specific type of traffic trunk. There also another approach is proposed in [22]. In this approach author proposed solution to prevent frequent preemption of low priority LSPs by high priority LSPs. Instead of preempting a low priority LSP, rate for this LSP is lowered to satisfy the request of new high priority LSP. This approach reduces the preemption rate of low priority LSP and thus reducing the network instability. Survivable online traffic engineering in MPLS networks has also been considered. Similar to MIRA, this scheme constructs LSPs dynamically by applying the shortest path algorithm to the dedicated link weight metric that reflects the specific traffic engineering requirement. This type of dynamic link metric is based on a Lost Flow in Link (LFL) function that is used to assign working routes with local restoration. In LFL the metric of a particular link reflects the change in the objective function if an incremental demand has been (re)routed through or even near that particular link [5].

3.4 Intra-Domain IP Based Traffic Engineering

Simple IP-based traffic engineering solutions have also been developed along with MPLS-based methods. In the IP-based traffic engineering solutions researchers tried to provide QoS guarantees by traditional hop-by-hop routing, without depending on MPLS based traffic engineering, using existing routing infrastructure. In [18], Ericsson et al. come up with a genetic algorithm-based method for the same IP-based traffic engineering optimization problem. In [19], Sridharan et al. propose a scheme based on the manipulation of a subset of next hops for some routing prefixes. The scheme is capable of achieving near-optimal traffic distribution without any change of existing routing protocols and forwarding mechanisms. Three different heuristic algorithms are used to optimally configure the next-hop of unicast destination prefixes. This method exhibits a typical strategy of making graceful trade-off between the performance and the overhead associated with the additional configuration needed. Wang et al. [20] propose Edge-Based Link Weight Setting, a new OSPF traffic engineering method without the necessity of ECMP division. Their method is to divide the physical network into several logical routing planes, each being associated with a dedicated link weight configuration.



3.5 Online IP-Based Traffic Engineering

Unlike offline traffic engineering, which has been extensively studied, there are also few proposals for online or adaptive IP-based traffic engineering. Two online traffic engineering methods are to change link weights on the fly and to make link weights sensitive to some loading or QoS parameters (e.g., to make the link weight a function of link utilization or delay). However, these methods require the flooding of new link weights throughout the network, which can cause route instability and looping problems during the convergence process. Another online traffic engineering method is to dynamically adjust the traffic division ratio according to the network load. Adaptive multipath (AMP) considers multiple nonequal cost paths and balances load by optimizing the traffic division ratios at each router. However, AMP only keeps network available information to a local scope rather than employing a global perspective of the network in each node [5]. 4.0 A PROPOSED SOLUTION Different ways have been introduced over time to address each of the above problems. DiffServ was introduced to classify IP traffic into multiple classes on the bases of its QoS requirements. MPLS traffic engineering has been used to gain the knowledge of network state and utilize this control the traffic admission and routing in the IP network. If we use DiffServ and MPLS traffic engineering together, QoS assurance can be enhanced for multimedia IP traffic. A conceptual diagram of the proposed network architecture is given as Figure 1. As can be seen from the figure, IP traffic is admitted at the edge routers and then routed via core routers to the destination edge routers. Traffic is differentiated into multiple classes according to QoS requirement of each class at the edge router. Edge routers also posses the network state knowledge about all links and possible paths from source to destination. Different paths are capable of satisfying different QoS parameters, e.g., a few paths will be able to satisfy bandwidth requirement, while others satisfy delay and jitter requirements. Having this knowledge and differentiated IP traffic, we can come up with an algorithm to satisfy different QoS requirements of differentiated traffic. Existing algorithms presented by other researchers rely on the bandwidth distribution efficiently among multiple traffic classes, somewhat ignoring the direct addressing of the delay and jitter requirements. Therefore our proposed solution will account for delay, jitter as well as bandwidth QoS parameters in the IP Core network. We will use existing knowledge of DiffServ and MPLS TE to improve existing QoS algorithm.

Fig. 1: IP/MPLS Core Supporting DiffServ



4.0 RESEARCH METHODOLOGY

Our work is based on but not limited to the development of QoS enhancement algorithms and then testing them with the help of simulation. Different QoS assurance strategies/algorithms will also be studied and a practical lab simulation model will be developed to demonstrate their shortcomings. In this simulation model, multiservice IP traffic will be injected to see the network and IP traffic behavior. An improved algorithm will be proposed, addressing the shortcomings in the existing algorithms. This new algorithm for each research issue will be simulated. Our new model can be tested in, but not limited to these scenarios. QoS assurance during normal network conditions using IP/MPLS TE. QoS assurance during heavily loaded network conditions without prioritizing the IP/MPLS traffic. QoS assurance during heavily loaded network conditions with prioritizing the IP/MPLS Traffic. Randomly fluctuating network topology. Creating short term network (link/system) failures. Creating long term network (link/system) failures. 5.0 CONCLUSION AND FUTURE WORK Increasing dependency of communication services on the internet packet based backbone requires QoS guarantees. Using internet (IP/MPLS) network infrastructure for voice, video and data services provide tremendous benefits in terms of cost and global presence. There are many challenges for the researchers to ensure QoS guarantees. In this study, existing QoS models will be enhanced to support future telecommunication services. Moreover, Network nodes, links, etc., are still vulnerable to under/over utilization, if traffic is not properly routed through the network. More seriously, failures cause sudden traffic shifts. The possibility of resource wastage in some network parts and over utilization in others is common. As the outcome of our study, we shall also introduce a robust routing model to enhance resource utilization and optimizing network throughput. 6.0 REFERENCES [1] D. Zhang, Ionescu “QoS Performance Analysis in Deployment of DiffServ-aware MPLS Traffic Engineering,”

Eighth ACIS International Conference, Volume 3, pages 963-967, July 30 2007-Aug. 1 2007 [2] Z. Li, Z. Zhang, L. Wang “A novel QoS routing scheme for MPLS traffic engineering,” Communication Technology

Proceedings, 2003. ICCT 2003. International Conference, Volume 1, 9-11 April 2003 [3] T. Kim, J. Yoo, H. Jung, H. Lee, S. Jin “Signaling for traffic engineering with QoS guarantee in IP environment,”

Advanced Communication Technology, 2005, ICACT 2005. The 7th International Conference [4] N. Hu, L. Li, Z.M. Mao, P. Steenkiste, J. Wang “Locating Internet Bottlenecks: Algorithms, Measurements and

Implications,” Proc ACM SIGOMM, 2004, pp.41-54 [5] N. Wang, K. Ho, G. Pavlou, M. Howarth “An Overview of Routing Optimization for Internet Traffic Engineering”

IEEE Communications Surveys 1st Quarter 2008, Volume 10, No.1 [6] L. Andersson, P. Doolan, N. Feldman “LDP Specification,” IETF RFC 3036, Jan. 2001. [7] D. Mitra, K. G. Ramakrishnan “A Case Study of Multiservice, Multipriority Traffic Engineering Design for Data

Networks,”Proc. IEEE GLOBECOM, 1999, pp. 1077–83. [8] O. Younis, S. Fahmy “Constraint-Based Routing in the Internet: Basic Principles and Recent Research,” IEEE

Commun. Surveys & Tutorials, 3rd qtr., 2003, pp. 2–13. [9] X. Xiao, A. Hannan, B. Bailey, L.M. Ni “Traffic Engineering with MPLS in the Internet,” IEEE Network, vol. 14, no.

12, Mar./Apr. 2000, pp. 28–33 [10] Z. Wang, J. Crowcroft “Quality of Service Routing for Supporting Multimedia Applications,” IEEE JSAC, vol. 14,

no. 7, Sept. 1996, pp. 1228–34. [11] R.A. Guerin, A. Orda, D. Williams “QoS Routing Mechanisms and OSPF Extensions,” Proc. IEEE GLOBECOM

1997, pp. 1903–08. [12] Y. Wang, Z. Wang “Explicit Routing Algorithms for Internet Traffic Engineering,” Proc. IEEE ICCCN, 1999, pp.

582–88. [13] F.L. Faucheur, W. Lai “Requirements for Support of Differentiated Services-aware MPLS Traffic Engineering,”

IETF RFC 3564, July 2003.



[14] P. Trimintzios, T. Baugé, G. Pavlou, P. Flegkas, R. Egan “Quality of Service Provisioning through Traffic

Engineering with Applicability to IP Based Production Networks,” Comp. Commun., vol. 26, no. 8, May 2003, pp. 845–60.

[15] V. Tabatabaee, B. Bhattacharjee, R.J. La, M.A. Shayman “Differentiated Traffic Engineering for QoS

Provisioning,” Proc. IEEE INFOCOM, 2005, pp. 2349–59. [16] A. Elwalid, C. Jin, S. Low, I. Widjaja “MATE: MPLS Adaptive Traffic Engineering,” Proc. IEEE INFOCOM, 2001,

pp.1300–09. [17] R. Boutaba, W. Szeto, Y. Iraqi “DORA: Efficient Routing for MPLS Traffic Engineering,” J. Network and Sys.

Mgmt., vol. 10, no. 3, Sept. 2002, pp. 309–25. [18] M. Ericsson, M.G.C. Resende, P.M. Pardalos “A Genetic Algorithm for the Weight Setting Problem in OSPF

Routing,” J. Combinatorial optimization, vol. 6, no. 3, Sept. 2002, pp. 299–333. [19] A. Sridharan, R. Guerin, C. Diot “Achieving Near-Optimal Traffic Engineering Solutions for Current OSPF/IS-IS

Networks,” IEEE/ACM Trans. Networking, vol. 13, no. 2, Apr. 2005, pp. 234–47. [20] J. Wang, Y. Yang, L. Xiao, K. Nahrstedt “Edge Based Traffic Engineering for OSPF Networks,” Comp. Networks,

vol. 48, no. 4, July 2005, pp. 605–25 [21] I. Iliadis, D. Bauer “A New Class of Online Minimum-Interference Routing Algorithms,” Proc. Second International

IFIP-TC6 Networking Conference, Pisa, Italy, May 19-24, 2002. [22] J.C.D. Oliveira, C. Scoglio, I.F. Akyildiz, G. Uhl “New Preemption Policies for DiffServ-Aware Traffic Engineering

to minimize Rerouting in MPLS Networks,” IEEE/ACM transactions on networking, vol. 12, no. 4, August 2004

7.0 BIOGRAPHY Muhammad Tanvir obtained his Master of Computer Science from Bahria University Pakistan in 2005. He worked at National Telecommunication Corporation, Data Communication Dept., Pakistan, from 2005 to 2009, He is presently a PhD student at Universiti Teknologi PETRONAS. His research area is QoS in Next Generation IP and MPLS networks. He holds many industry vendor certifications on IP Networks. Abas Md Said obtained his PhD from Loughborough University, UK. He is an associate professor at Universiti Teknologi PETRONAS. His area of research includes computer graphics, visualizations and computer networks.



Proceeding of the 3rd International Conference on Informatics and Technology,

Documents

internet traffic engineering

qos models

internet backbone

traffic engineering

qosguaranteed services

internet service providers

required qos

endtoend qos requirements