Optimizing MPLS Traffic Engineering - Assignment Example

Proposals for Optimising MPLS Traffic Engineering Abstract An important effect of traffic engineering is the elimination of path congestion. TE does not necessarily go for the shortest distance between two domains; two different packets can travel over two different paths even though they may come from the same node and end up in the same node. In this way less used network resources can be used. Internet Traffic Engineering is the area of Internet network engineering concerned with the performance optimization of traffic handling in operational networks, with the intention to reduce over-utilization of capacity when other capacity is available in the network. In particular it relates to the design, provision and tuning of operational internet networks.  It applies business goals, technology and scientific principles to the measurement, modelling, categorization and control of internet traffic, and use of such information for service and performance objectives, including the reliable and expeditious movement of traffic through the network, the efficient utilization of network resources, and the planning of network capacity. This essay will explore options available to optimize TE under MPLS techniques. Introduction There are various techniques and approaches that can be used to manage traffic in operational networks. Some of the techniques that may be employed include ATM and Frame Relay overlay models, MPLS based approaches, constraint-based routing, and traffic engineering methodologies in Diff-Serv environments, LSP models, Algorithms, and intra-domain and inter- domain traffic engineering. Traffic Engineering According to Awduche et al. 1999, “A major goal of internet traffic engineering is to facilitate efficient and reliable network operations while simultaneously optimizing network resource utilization and traffic performance .Traffic engineering has become an indispensable function in many large autonomous systems because of the high cost of network assets and the commercial and competitive nature of the Internet. These factors emphasize the need for maximal operational efficiency”. There are two performance objectives to be observed in traffic engineering; traffic and resource oriented. Traffic oriented performance objectives are generally targeted at Qos results which can be reduction of packet loss, delay, maximization of output, and meeting of service level standards. The reduction of packet loss remains one of the main objectives of traffic oriented performance. Resource oriented performance objectives relate to the optimal use of resources. The efficient use of network resources is a means for the attainment of resource oriented performance objectives. It is better avoid the over-utilisation and congestion of certain network resources while other subsets along alternate feasible paths that remain underutilized exist. Because bandwidth is a crucial resource in modern networks; for this reason efficient Traffic Engineering should efficiently manage bandwidth resources. Traffic engineering is aptly suited to the solving of congestion problems, especially those resulting from inefficient resource allocation. This kind of congestion can be reduced by adopting mechanisms that balance loads over the network. The objective is either to reduce maximum congestion or to reduce maximum resource allocation by making efficient resource allocation. Congestion reduction leads to a reduction in packet loss, transit delay and aggregate throughput increases. This translates to a better service for the clients and or users. Whatever policies are chosen for traffic engineering they should be flexible enough so that other policies can be implemented in consideration of current cost structure and the utilisation of the service. Essentially a traffic engineering system is composed “of interconnected network elements, a network performance monitoring system and a set of network configuration management tools. The Traffic Engineer formulates a control policy, observes the state of the network through the monitoring system, characterizes the traffic, and applies control actions to drive the network to a desired state, in accordance with the control policy” (Awduche et al, 1999). A traffic engineer monitors the system (the state of the traffic), characterizes the traffic and applies appropriate controls as formulated. Control actions can involve three actions; modification of traffic management parameters, modification of parameters associated with routing, and reorganisation of properties and restrictions of the resources. The overlay model is a popular choice for expanding the design space by enabling discretionary virtual topologies to ride over a network’s physical topology. For example using IP over ATM PVCs. However it is hard to manage because it requires configuration at each router in the network because it works with two protocols. MPLS allows for the use of only one protocol. MPLS: MultiProtocol Label Switching The importance of traffic engineering relates to its strategic capability to provision the same functionality offered by the overlay model in an integrated way and at less cost. It also offers automation capabilities. Multiprotocol Label Switching (MPLS) essentially combines label switching techniques with routing techniques. “The basic idea involves assigning short fixed length labels to packets at the ingress to an MPLS cloud (based on the concept of forwarding equivalence classes). Throughout the interior of the MPLS domain, the labels attached to packets are used to make forwarding decisions (usually without recourse to the original packet headers” (Ibid). MPLS has certain advantages which are: (1) explicit label switched paths which are not constrained by the destination based forwarding paradigm can be easily created through manual administrative action or through automated action by the underlying protocols, (2) LSPs can potentially be efficiently maintained, (3) traffic trunks can be instantiated and mapped onto LSPs, (4) a set of attributes can be associated with traffic trunks which modulate their behavioral characteristics, (5) a set of attributes can be associated with resources which constrain the placement of LSPs and traffic trunks across them, (6) MPLS allows for both traffic aggregation and disaggregation whereas classical destination only based IP forwarding permits only aggregation, (7) it is relatively easy to integrate a “constraint-based routing" framework with MPLS (8) a good implementation of MPLS can offer significantly lower overhead than competing alternatives for Traffic Engineering ( Awduche et al, 1999.). Awduche et al. 1999 identified three MPLS problems; 1. how to copy packets onto forwarding equivalence classes 2. how to copy forwarding equivalence classes onto traffic trunks. 3. how to copy traffic trunks onto the physical network frame through LSPs. This means that it is necessary to undertake optimization for MPLS to fully actualize policies that will lead large operational networks to the optimal performance. This can be done using any of the models below. In MPLS LSPs or Label Switching Protocols are created with path instructions according to user specified requested policies. Signalling protocols: RSVP-TE and CR-LDP These two LDPs support explicit and constraint based routing. MPLS traffic engineering allows for the use of a single protocol but to do this it needs a label distribution protocol that supports constraint and explicit based routing like RSVP and CR-LDP. CRD-LDP: It is a set of LDP extensions which are meant to facilitate constraint based routing of LSPs. It employs TCP sessions to relay messages along the sessions; allowing for the reliable distribution of control messages. It makes an attempt to make LDP protocol to function over explicit route carrying traffic parameters for resource reservation and options for CR-LSP strength features. Messages are sent once, there is no need for refreshing –information. Before an explicit route can be established a label request message with a list of nodes on the intended constraint-based route is sent. The signalling message travels to the destination over the selected path, and if the requested path is able to satisfy the requirements, labels are allocated and distributed by means of LABEL MAPPING messages starting with the destination and propagating in the reverse direction back to the source. Assuming that resources are available, the LSP setup is completed after a single round-trip of the signalling message. CR-LDP is capable of establishing both strict and loose path setups with setup and holding priority, path pre-emption, and path re-optimization. CR-LDP enables multiprotocol operations by using an opaque FEC, which allows core LSRs to ignore the type of traffic being transported across the network. The opaque FECs are also used for security purposes as well, i.e. not enabling the LSRs to know the transport data services identity (Kim & Chung). RSVP-TE: RSVP-TE Resource Reservation Protocol- Traffic Extensions is an extension of basic RSVP. RSVP is a resource reservation protocol that ensures quality integrated service over internet. RSVP is used by a host or routers to send out a message for resource allocation; to request specific qualities of service from the network or to deliver quality-of-service (QoS) requests to all nodes along the path(s) of the flows and to establish and maintain state to provide the requested service respectively. Because of the soft state mechanism of RSVP (using PATH and RESV commands to establish a LSP) problems can occur in resources allocation. RSVP-TE is a receiver-oriented protocol, meaning that label allocation and bandwidth reservation are driven by the receiver node. In particular, the standard specifies that label allocation has to be executed in downstream on-demand mode that is the label is created by the downstream node and distributed to the previous hop only. RSVP-TE as an extension of RSVP supports ER-LSPs and many modifications and extensions support coping with traffic engineering needs. These extensions mainly support addition of traffic engineering capabilities and resolution of scalability problems. RSVP-TE and CR-LDP compared A number of differences mark the RSVP-TE. Here is list of differences and similarities. CR-LDP support RSVP Support Transport TCP Raw IP Security Yes1 Yes1 Multipoint-to-Point Yes Yes Multicast Support No2 No2 LSP Merging Yes3 Yes3 LSP State Hard Soft LSP Refresh Not needed Periodic, hop-by-hop High Availability No Yes Re-routing Yes Yes Explicit Routing Strict and loose Strict and loose Route Pinning Yes Yes, by recording path LSP Pre-emption Yes, priority based Yes, priority based LSP Protection Yes Yes Shared Reservations No Yes Traffic Parm Exchange Yes Yes Traffic Control Forward Path Reverse Path Policy Control Implicit Explicit Layer 3 Protocol Indicated No Yes Resource Class Constraint Yes No Source: Mpls Traffic Engineering: A Choice Of Signalling Protocols, Paul Brittain and Adrian Farrel, 2000 DiffServ-aware MPLS traffic engineering It allows different traffic classes to take different routes through the network. Unlike DiffServ, DS-TE is a control plane feature, i.e. its effect is achieved through signalling and not through data plane processes such as queuing (Smith ).To do this the bandwidth is segmented into different tunnels while maintaining existing pre-emption rules. Further modifications are made to the CSPF constraint shortest path first algorithm. According to Smith, Path computation makes use of revised bandwidth segmentation rules to route different tunnels by different paths across the network. For example tunnel for best effort traffic and one high priority voice traffic can be separated. Once the tunnel is set up, the headend router can allocate the traffic to a particular tunnel, based on static routing or dynamic routing, using the tunnel virtual interface. A best-effort-based traffic-engineered network finds the appropriate route for traffic and sets up the aggregate tunnel. In a traffic-engineered, while in a DiffServ environment, a tunnel is set up for each class the approach is to set up a tunnel for each class. For example, to achieve voice trunking in a multiservice network, DiffServ-aware MPLS-TE can be used to provide explicit admission control of EF traffic/voice trunks using EF-aware constraint-based routing. Traffic engineering and DiffServ-aware MPLS combined can provide high QoS for voice without relying too much on engineering. LSP Models LSP Tunnel: A tunnel that tunnels below a normal IP or filtering mechanism, an LSP tunnel allows for the application of network optimisation processes. These optimisation processes like avoiding network failures, congestion and bottlenecks can be applied to LSP tunnels either automated or manually. Furthermore, they can be established between two nodes, and traffic between the two nodes can be transferred onto the LSP tunnels according to requirements. AN LSP tunnel is created by having the sender node sends an RSVP Path message with session type ipv4 or ipv6; with a label request object into the path message. The label request object outlines a request for label binding for the chosen path and also provides an indication of the network layer protocol to be used on the path; because the network layer protocol travelling along an LSP cannot be read as necessarily as IP because it cannot be inferred from the L2 header. The L2 header only identifies the higher layer protocol as MPLS. A route can be chosen for all or part of a session due to any of these reasons; if it meets the tunnel’s Qos indications, is efficient with resource use or meets any other criteria. To do this; the originating node adds an explicit route object to the RSVP path message. The explicit route object outlines the route as a series of abstract nodes. LSP Hose model In order for a private network to provide security and privacy over an existing physical network like the internet; while saving on the cost of physical link installation it needs various techniques as IPsec, tunnelling, and special routing protocols.. The hose model is both a VPN service interface and also a performance abstraction. It simplifies specification aspects of a VPN. A hose model ensures bandwidth limits and, while it provides a link into the network, it allows sending and receiving traffic without the need of predicting point to point loads. The pipe model on the other hand allocates resources and manages through the replication of private lines from one VPN terminal to all other terminals. LSP Funnel (multipoint-to-point): The funnel model which is a scalable model that groups resource requests using Network Resource Management NRMs. NRMS are capable of making resource requests from other NRMs. A number of resource requests heading to the same point are grouped, use the same path in such a way that each NRM has one reservation per destination domain. The destination NRM will disperse the resource requests it receives to all the endpoints within its jurisdiction. One weak point of this model is that it assumes that destination domain will be well provisioned to meet Qos indications which are not always the case especially with very large networks. Therefore it cannot be used for end-to-end Qos if the destination is under provisioned. Also destinations like wireless connections are not well provisioned and are not suitable for use with this model. In mobile wireless networks, Qos levels should be fulfilled by ensuring quick local re-computation of qualities between base stations in the wireless network. Inter –domain and Intra-domain TE Intra-domain: In an AS two TE techniques can be used; pure IP networks and those that rely on the use of Multi-Protocol Label Switching (MPLS). Intra-domain routing protocol, or Interior Gateway Protocol – IGP can be used in a pure IP network, to control the flow of the IP packets. Inside a domain, the routers compute the best path to reach each destination in order to populate their forwarding table. The best path is usually the path with the smallest cost where the cost of a path is the sum of the cost of all the links that compose the path. The cost of the links is distributed by the routing protocol. The cost associated with each link depends on the links or paths that the Internet Service Provider (ISP) wants to favour. If each link has a unitary cost, then the routing protocol favours paths with the smallest number of hops. If the cost of each link is a function of the link transmission delay, the routing protocol selects paths with the shortest delay. If the cost associated with a link is a function of the link bandwidth, the routing protocol favours high bandwidth paths. To have robust QoS features and control over the domain, it is necessary to implement MPLS technology over Diff-serv domain. For end-to-end QoS in an MPLS network it is important to solve the inter-domain Label Switched Path (LSP) setup problem. There are different approaches for interdomain label distribution. One of the approaches is to use Border Gateway Protocol (BGP) to distribute the label between peer edge LSRs. In this approach the label information is transmitted with the BGP update message. This is done by using the BGP-4 Multiprotocol Extensions attribute. Inter-domain: Inter-domain traffic engineering techniques are concerned with the engineering between Autonomous systems, AS. They rely mostly on the Border Gateway Protocol (BGP). They are employed by for example internet service providers as they do employ inter-domain TE techniques. Inter-domain TE is main purpose is basically the optimized performance for traffic from one domain to another. The BGP traffic engineering techniques aim at modifying the BGP routes selected by the routers in order to change the paths followed by the traffic. The BGP TE techniques aim at modifying the BGP routes selected by the routers in order to change the paths followed by the traffic. A router that receives several BGP routes for a prefix selects one of these routes by means of a decision process based on any of these options Prefer routes with highest LOCAL_PREF, 2 Prefer routes with shortest AS-PATH, 3 Prefer routes with lowest MED, 4 Prefer eBGP over iBGP routes, 5 Prefer routes with lowest IGP cost to the next-hop, and 6 Tie breaking rules. By applying any these options, Inter-domain TE can avoid redistributing certain routes or change the values of certain attributes to favour or penalize certain routes. Intra-domain and inter-domain TE: Inter-domain TE is responsible for inter-domain routing. For example by implementing and a Qos enhanced BGP. Inter-domain TE also dynamically performs load balancing between the multiple paths defined by Offline Inter-domain TE. It uses real-time monitoring information, changing appropriately the ratio of the traffic mapped to the inter-domain paths. Intra-domain TE includes routing, load balancing and dynamic bandwidth assignment for managing in real-time the resources allocated by Offline Intra-domain TE, in order to react to statistical traffic fluctuations and other conditions. Conclusion: This essay has outlined a number of approaches in which MPLS can be optimized. Any of the approaches can be chosen depending on the context and the size of the network. With the ever growing need of connectivity between networks and the need for efficiency these approaches can only grow in importance and their improvement more important. Because of its applicability to large MPLS is indispensable for internet networking. The internet growth is set to grow steadily so MPLS is going to be useful in the future. Conventional routers have been unable the kind of internet traffic of the last 15 years. MPLS can work over different routing protocols, transport layers and addressing schemes. Bibliography 1. Smith, Ed.2007, Comparison of Traffic Engineering Techniques in Connectionless and Connection-Oriented Networks, The Journal of The Communications Network , Volume 6 Part 2 , April–June 2007 2. Kim, Sang-Chul, and Chung Jong- Moon, 2007, Analysis of MPLS Signaling Protocols and Traffic Dissemination in OSPF and MPLS, Department of Computer Science, Kookmin University School of Electrical Engineering, Yonsei University 3. Chung Jong-Moon, 2000, Analysis of MPLS Traffic Engineering, IEE Director of the Advanced Communication Systems Engineering Laboratory School of Electrical and Computer Engineering, Oklahoma State University, engineering South, Stillwater, OK 74078, U.S.A. 4. Pelsser, C. 2006, Interdomain Traffic Engineering with MPLS by Thesis submitted in partial fulfillment of the requirements for the Degree of Doctor in Applied Sciences, Faculté des Sciences Appliquées Département d’Ingénierie Informatique Université catholique de Louvain, Louvain-la-Neuve Belgium 5. Ash,G. 2002,Traffic Engineering & QoS Methods for IP-, ATM &TDM Based Multiservice Networks, Available at: http://www.tools.ietf.org 6. WIPO (Online) 2003, Method and Arrangement in An IP Network, Available at:http://www.wipo.int/pctdb/en/wo.jsp?WO=2003%2F021888&IA=WO2003%2F021888&DISPLAY=DESC 7. Awduche, D. et al. 1999, Requirements for Traffic Engineering over MPLS, Available at http://www.apps.ietf.org/rfc/rfc2702.html 8. IEC (online)2007, Multiprotocol Label Switching (MPLS) , Available at: http://www.iec.org/online/tutorials/mpls/topic03.html 9. Brittain Paul, Farrel Adrian, 2000, MPLS Traffic Engineering: A Choice of Signalling Protocols, Analysis of the similarities and differences between the two primary MPLS label distribution protocols: RSVP and CR-LDP. Available at: http://www.dataconnection.com Read More