MPLS
(Multiprotocol Label switching)

Introduction

The rapid growth in the internet presents various challenges on current IP based carrier networks. New applications require services, which are deterministic in nature. The specific service characteristics required by the applications must be guaranteed a cross the complete path of the network in which the application data traverses. Current routing technology utilizes the best available path information based only on the destination address, however the application data's attributes are not considered. As the network grows there is an increased demand on the routers to handle huge amounts of routing information in addition to applications data. Besides, the forwarding decision made at each hop as a packet travels from one router hop to another inhibits sc alability and performance. Multi Protocol Label Switching(MPLS) is a technology, which addresses some of these issues.

        MPLS combines layer 2 (data link layer) switching technologies with layer 3 (network layer) routing technologies to form a new level of standard. The main goal of this standardization is to create a new terminology that provides increased performance and scalability. In this paper we highlight on the challenges faced by service providers, in moving to IP-based networks, the current capabilities of IP, the current methods used to build large-scale IP netw orks, and what will MPLS do in order to fulfill the requirements of the service provider community.

MPLS Background

Internet Engineering task Force (IETF) defines MPLS as an approach to applying label switching technology to large-scale networks. IETF is defining MPLS, as there are many issues related to it which need globalization. Among these issues are scalability o f IP networks to meet the growing demand, Combining different types of traffics into a single IP network and improving efficiency. MPLS as per its name can support multiple protocols, but the initial work is focused on the integration of IP4 with ATM and frame relay.

MPLS Overview

          
Multiprotocol label switching is a technique, which allows high-performance transport of IP traffic across Wide Area Networks. It is possible to set up explicit routes for data flows that are constrained by path, resource availability and requested Qualit y of Service (QOS) by using MPLS techniques. It also facilitates highly scalable Virtual Private Networks (VPNs). MPLS assigns labels to IP flows and hence placing them in IP frames. These frames can be transported across cell or packet based labels inste ad of being routed by IP address look-up.
 
        ATM is a very good protocol for handling any type of traffic that requires dedicated circuits, Quality of Service (QoS) and heavy-duty network management in layer 2. There were lot problems in mixing IP with ATM. IP uses packets of any size. ATM uses 53-byte cells. Placing IP packets into those ATM cells results in lot of bandwidth being wasted.

MPLS is a latest protocol, which is defined by Internet Engineering Task Force (IETF) that streamlines and simplifies IP addressing. Two things are done when a router checks an IP packet header. The router first assigns each packet is assigned to a for ward equivalence class (FEC). After that each FEC is assigned to a specific next hop. The entire process is repeated each time a packet reaches a new router. If MPLS is used, each packet is assigned to an FEC only once, which is at the time when it enters the network. The FEC is encoded as a value called a label and that label is sent with the packet at each hop. The label is used as an index to a lookup table at the node to get a new label when the packet encounters the next node in its path. The new lab el replaces the old label and the packet is sent. As the label also can contain that data, ATM's network management and QoS capabilities can be replaced by MPLS.
      
Nowadays faster and higher bandwidth ATM switches are being out-performed by Internet backbone routers. MPLS offers simple mechanism for packet-oriented and multi-service functionality with the added benefits of greater scalability.

As promising as it seems, MPLS isn't yet a proven technology. Efforts are going on to make MPLS support our traffic engineering at least as well as ATM.

MPLS has some key terminologies and concepts like:

· Datagram: Unit of information sent through a network.
· LSP (Label switched path): It is the path that a datagram travels through a network based upon the labels that are assigned to that datagram.
· Ingress: the first label switch in LSP
· Egress: the last label switch in LSP
· Label Switching: It is the general technology that combines layer 2 and layer 3 technologies.
· LSR (Label switching router): It is a device that is in the middle of the network and is capable of forwarding datagrams based upon a label.
· MPLS is the term used to describe the specific work going on in the IETF to standardize label switching.
· MPLS Domain is a portion of network that contains devices that understand MPLS.

Benefits of MPLS

Initially Internet was designed to serve limited community, where services like e-mail, file transfer and remote logon played a prominent role. With the introduction of personnel computers and World Wide Web (WWW) the use of internet has reached its limit s, which in turn has forced the networks to increase in there network ability to handle rapidly increasing traffic. With the introduction of newer technologies like Video conferencing the demand of differentiation of service offering have changed a lot.
        
Under present situations Internet cannot support these requirements, instead it offers no guarantees, given best-effort service with software-based network–layer routing.

The Multiprotocol Label Switching technology addresses these requirements.

1. MPLS protocol supports multipath routing and forwarding.
2. Scalability of network layer routing i.e. using labels as a means to aggregate forwarding information, while staying in routing hierarchies.
3. MPLS forwarding allows stream of data to be forwarded as a unit and shows that an identified stream takes a single path.
4. Using labels to identify particular traffic, which are to receive special service e.g. QOS we can achieve greater flexibility in delivering routing services.
5. MPLS switches can work with non-MPLS switches in the same network.
6. Increase price-per-performance by using label-swapping paradigm to optimize network performance.
7. Simplify integration of routers with cell switching based technologies.

a. Making cell switches behave as peers to routers, reducing the number of routing peers that a router has to maintain.
b. Making available the information of physical topology to network layer routing procedures
c. Using common addressing, routing and management procedures.

What MPLS Does:

Latest technology includes Internet backbone routers, new queuing and scheduling algorithm, integrated routing/forwarding solution etc. These all technologies are critical for the growth of Internet and for its successful operation. MPLS is the latest adv ancement in the evolution of routing/forwarding technology for the core of Internet. MPLS offers a solution that integrates the control of IP routing with the simplicity of layer 2 switching.

Role of Routing:

Packet forwarding is based on datagram routing i.e. IP packet is routed through the network based upon the destination address contained within the packet. Here the terminology followed is Hop by Hop i.e. every packet entering a router is examined and a decision is made as to where to send the packet. This way packet moves from source to destination “Hop by Hop”. These types of networks are called connectionless networks. A router may have more than one adjacent routers; hence there should be some way to figure out the next hop of the packet. For this we may use any routing protocol (e.g. open shortest path first). Now using the information passed by the routing protocol, the routers build forwarding tables that identify the appropriate next hop for all known IP destination addresses.

Role of Switching:

As Large IP networks became a necessity; these router-based networks had a number of issues to be taken care of. Some of the issues related to these types of networks were cost of high-speed routers, efficiency and reliability of routers in large netwo rks.

      Hence a new technology i.e. Switching came in picture, which was based on ATM and frame relay. This utilizes a much different forwarding algorithm called label-swapping algorithm. Since this algorithm is simple, it i s typically implemented in hardware, yielding a better price/performance as compared to IP routing. This terminology is connection oriented i.e. traffic travel between two end points once a connection is established. Thus these types of networks are more manageable and predictable.

Merging Routing and Switching:

Switching has an upper hand as compared to IP Routing but still IP routing dominates the edge of the network. Due to the need of large networks it became necessary to interconnect these two technologies. The IP has been overlaid on top of ATM or Frame relay. Since paths between routers now go through switches, they require connection thus increasing the reliability of networks. When IP is overlaid on ATM all of the routers appear to be directly connected to each other at the network layer. Since all ro uters now should be connected directly or indirectly, the network now requires a mesh of Virtual Circuits (VCs) to interconnect the routers. As the routers grow the number of required VCs grow at the rate of n(n-1)/2. As network grows the number of router s increases, the VC requirement grows exponentially, thus limiting the scalability of overall network.

Advantages of MPLS over Conventional IP-over-ATM:

MPLS offers simpler integration of control information, IP routing protocol run directly on the label switches, the topology of the ATM is visible to IP routing, making routing at IP layer more likely to be optimal. Scalability of IP routing is improve d.

MPLS solves a number of complex problems because it provides a base that supports the successful implementation of advance routing services:

MPLS deals with the scalability issues associated with current IP-OVER ATM model, MPLS reduces the complexity of network operation, MPLS enhances the new routing capabilities that will help improve the current IP routing techniques, MPLS offers standar d solution, which can be applied on any platform.

Basic Components of MPLS:

What is a Label?

Label is a short is a short fixed length, which is physically contiguous, and is a locally significant identifier used to represent the Forwarding equivalence class to which the packet is assigned.

FEC: (Forwarding Equivalence Class) is an important concept in MPLS. An FEC is any subset of packets that are treated the same way by a router. By “treated” this can mean forwarded out the same interface with the same next hop and label. It c an also mean given the class of service, output on same queue, given same drop reference, and any other option available to the network operator. When a packet enters the MPLS network at the ingress node, the Packet is mapped into an FEC. The mapping can also be done on wide variety of parameters, address prefix, source destination address pair, or ingress interface. This greater flexibility adds functionality to MPLS that is not available in traditional IP routing.

A label has only the information about the path that has to be followed it doesn't encode information from the network layer header, with that information it is used to map traffic to a specific FEC.

Label Distribution Protocol (LDP)

The distribution of labels in MPLS is accomplished in one or more ways.

· Extending routing protocols such as Border Gateway Protocol (BGP) to support label distribution.
· Using the resource ReSerVation Protocol (RSVP) signaling mechanism to distribute labels mapped to the RSVP flows.
· Using the Label Distribution Protocol (LDP) as defined by the IETF.

The Label Distribution Protocol (LDP) is a set of procedures and messages by which LSRs establish LSPs through a network by mapping network-layer routing information directly to data-link layer switched paths. These LSPs may have an endpoint at a direc tly attached neighbor (comparable to IP hop-by-hop forwarding), or may have an endpoint at a network egress node, enabling switching via all intermediary nodes.

LDP associates an FEC with each LSP it creates. The FEC associated with an LSP specifies which packets are “mapped” to that LSP. LSPs are extended through a network as each LSR “splices” incoming labels for an FEC to the outgoing la bel assigned to the next hop for the given FEC. Discovery messages provide a mechanism whereby LSRs indicate their presence in a network by sending the Hello message periodically. This is transmitted as a UDP packet to the LDP port at the 'all routers on this subnet' group multicast address. When an LSR chooses to establish a session with another LSR learned via the Hello message, it uses the LDP initialization procedure, the two LSRs are LDP peers, and may exchange advertisement messages. The LSR request s a label mapping from a neighboring LSR when it needs one, and advertises a label mapping to a neighboring LSR when it wishes the neighbor to use a label. The LDP uses the transmission Control Protocol (TCP) transport for session, advertisement and notif ication messages as the TCP provides reliable and in order delivery of messages.
Ref #14

In the MPLS architecture, the decision to bind a particular label L to a particular FEC F is made by the LSR, which is DOWNSTREAM with respect to that binding. The downstream LSR then informs the upstream LSR of the binding. Thus labels are “downs tream-assigned”, and label bindings are distributed in the “downstream to upstream” direction. If an LSR has been designed so that it can only look up labels that fall into a certain numeric range, then it merely needs to ensure that it onl y binds labels that are in that range.

The MPLS architecture allows an LSR to explicit request, from its next hop for a particular FEC, a label binding for that FEC. This is known as “downstream-on-demand” label distribution. The MPLS architecture also allows an LSR to distribute bindings to LSRs that have not explicitly requested them. This is known as “unsolicited downstream” label distribution. It is expected that some MPLS implementations will provide only downstream-on-demand label distribution, and some will provid e only unsolicited downstream label distribution, and some will provide both. Which is supported may depend on the characteristics of the interfaces, which are supported by a particular implementation. However, both of these label distribution techniques may be used in the same network at the same time. On any given label distribution adjacency, the upstream LSR and the downstream LSR must agree on which technique is to be used.

An LSR Ru may receive (or have received) a label binding for a particular FEC from an LSR Rd, even though Rd is not Ru's next hop (or is no longer Ru's next hop) for that FEC. Ru then has the choice of whether to keep track of such bindings, or whether to discard such bindings. If Ru keeps track of such bindings, then it may immediately begin using the binding again if Rd eventually becomes its next hop for the FEC in question. If Ru discards such bindings, then if Rd later becomes the next hop, the bi nding will have to be reacquired. If an LSR supports “Liberal Label Retention Mode”, it maintains the bindings between a label and a FEC, which are received from LSRs, which are not its next hop for that FEC. If an LSR supports “Conservativ e Label Retention Mode”, it discards such bindings. Liberal label retention mode allows for quicker adaptation to routing changes, but conservative label retention mode though requires an LSR to maintain many fewer labels.

The main requirement in MPLS is that a LSR forwarding label switched traffic to another LSR, puts a label on that traffic which is understood by other LSR. There are various ways by which a LSR's can learn about each other, they are:

1) Explicit label Distribution
2) Downstream Label Allocation
3) Upstream Label Allocation
4) Other Label Allocation Methods
   

Explicit Label Distribution:

         Suppose LSR A is forwarding traffic to LSR B. A is called upstream LSR and B the downstream LSR. A applies a Label on the traffic such that B understands. One thing should be taken into consideratio n, that is the meaning of label should be communicated between A and B. Here A and B are in the same domain which can be widely spread. Various approaches are followed for explicit label distribution.

Downstream Label Allocation:

        In this method label allocation is done by downstream LSR, i.e. the LSR that uses the label as an index to look into the switching table. This method is basically followed in unicast traffic. Here we use control driven model (which is discussed in later part of the paper) for allocation of labels. Label allocation should be done with care as subsequent forwarding decisions are made on that basis. Once the labels are allocated, they are then distributed to all the peers. For this a special type of distribution technique is used in which, a LSR requests a label for a route from particular peer only when its routing calculation indicate that peer to be the new next hop for the route.

Upstream Label Allocation:

        Here label allocation is done by upstream LSR. In this case LSR choosing the label and the LSR, which needs to interpret packets using the label, are not the same node. Label at issue is not used as an in dex into the switching tables but it is found as the result of a look up on those tables. In upstream assignments the same label is used on all output ports for which a particular packet was destined.

Other Methods

This includes using label values, which are unique within the MPLS domain. Hence, any traffic to a particular MPLS egress node could make use of the label of that node. One difficulty in this method is to find out the scope over which a label is unique . One way to solve this is to configure each node in an MPLS domain with a label, which is unique in that domain.
Ref#12

Fundamental building blocks common to all multiplayer switching solutions.

MPLS a multilayer solution can be divided into two distinct functional components.

1. Control Component
2. Forwarding Component

Control Component:

Information has to be stored about the manner in which packets will be routed.
The control component does this by building and maintaining a forwarding table, which will have the information. It also communicates with other routers using standard routing protocols.

Forwarding Component:

The manner in which packets will be forwarded and routed is determined by the forwarding component. The forwarding component does this job by looking up the forwarding table when a packet arrives. It maps the information in the packet header with the l ook table to determine the next and then directs it to the destination.

By using this mechanism in which we develop Control and Forwarding components separately, we get a benefit of developing and modifying each component separately. The Control component communicates with the Forwarding component by managing the packet-fo rwarding table.

Label swapping is used to implement the forwarding component.

Label swapping is used to implement the forwarding decision of a packet.
Let A and B be 2 Label switch routers.

Label Swapping (Forwarding component)

When a packet comes to LSR, A which is the first label switch (ingress), an initial label is assigned to each packet. It also receives the destination address where the packet is to be routed. The ingress switch maps the incoming packet to an FEC by perfo rming a match with the forwarding table. In terminology there exists a stack of labels called the NHLFE (Next Hop Label Forwarding Entry). It is this entry, which decides the next hop in forwarding a packet. The NHLFE is implemented as a FIFO stack. When a new label is to be assigned it pops out a label from the top of the stack and replaces it with the new label. A path is virtually defined between the ingress and the egress about the entire route. Thus when an appropriate match is found in the forwardin g table an outgoing label is retrieved which contains the information about the next hop. This outgoing label, which is popped out of the stack, is replaced with incoming label and hence we say that a label swap has occurred. Thus the forwarding component performs a label swap at each hop to determine the next hop and finally when it reaches the last label switch, the Egress and tries to perform a label swap and a next hop is not available, the incoming label is ignored and the packet is forwarded using I P forwarding.

Observations, Issues and Assumptions for MPLS

The common approach adopted by multiplayer switching solutions was to take the control software from an IP router, integrate it with forwarding performance of a label-swapping ATM switch, and create an extremely fast and cost efficient IP router.

Layer 2 versus layer 3:

In order to provide simple and fast packet forwarding capability, MPLS uses Layer 2 switching. A node, which must be forwarded in layer 3, must parse a large header, and to determine forwarding path a longest-prefix match has to be performed. Direct in dex lookup can be done into its forwarding table with the short header when MPLS label switching is performed and when the labels are assigned properly. As label-switching function is less complex, building a label switching hardware is simpler than it is to build a layer 3 forwarding hardware.

Scalability:

Scalability in MPLS is given by two principles of routing. The first principle is that forwarding follows an inverted tree rooted at a destination and the second principle is that routing aggregation reduces the number of destinations.

Types of Streams:

There are four different types of switched paths in MPLS. They are
 
· Point to point: Used to carry unicast traffic by connecting the ingress nodes to the egress nodes.
· Multipoint to point: Used to connect the ingress nodes to a single egress node.
· Point to multipoint: Used to distribute multicast traffic.
· Multipoint-to-multipoint: Used to take the multicast traffic, which comes from different sources into a single multicast distribution tree.

The factors that determine the type of switched path to be used are: the reason for creating the switched path and the efficiency of the switches in the network.

Point to point is used to carry unicast traffic by connecting the ingress nodes to the egress nodes. Multipoint to point connects the ingress nodes to a single egress node. Data driven versus control traffic driven label assignment is discussed in late r part of the paper.

Association of labels and network layer routing is a fundamental concept in MPLS. There are three ways in Label assignment:

· Topology driven label assignment: In this scheme labels are assigned in response to normal processing of routing protocol control traffic. In this assignment if path exists a label has been assigned to it. Bandwidth consumed by label distribu tion depends on the size of network; labels assigned by this assignment may travel through highly busy routes.

· Request driven label Assignment: In this scheme labels are assigned in response to normal processing of request based control traffic. Bandwidth consumed by label distribution depends on the amount of control traffic in the system. In this ass ignment also, if path exists a label has been assigned to it. Depending on the number of flows supported label size varies.

· Traffic driven label assignment: In this scheme arrival of data at LSR enables label assignment and distribution. In this case sort lived but recurring flows may impose a heavy control burden therefore it requires high performance packet class ification capabilities.

As mentioned earlier each multilayer switch ran standard IP routing software and a proprietary label-binding mechanism. The routing s/w permitted multilayer switches to exchange layer 3 network reachability information whereas label-binding mechanism m apped layer 3 routers to labels and distributed them to neighbors to establish LSPs across the core of the network.

Running routing protocol on core system rather than just edge systems provided a number of benefits that enhanced network operation:

Eliminates the IP over-ATM model's “n squared” PVC scaling problem, Reducing interior Gateway Protocol stress by dramatically decreasing the number of peers that each router had to maintain, Permitted information about the core's actual physi cal topology to be made available to network layer routing procedures.

In forwarding component, multilayer switches used conventional ATM switching hardware and label swapping to forward cell across the core of the network. Labels to routes were assigned by control procedures, which in turn distributed the labels among mu ltiplayer switches and created the forwarding tables, which were managed by proprietary IP-based protocol, not ATM forum protocols.

The benefits of ATM label swapping in the core of network are:

Label swapping improved network performance by leveling the benefits of hardware-based forwarding, It made explicit routing practical i.e. the path that traffic will follow is preconfigured which is different from the one typically found out by destina tion-based routing, Multiplayer switching provides an enhanced forwarding control than, the one supported by traditional routing mechanism.

Hence multilayer switching reduced operational complexity by eliminating the need to coordinate and map between two different protocol architectures IP and ATM. A critical limitation of multilayer switching solution was that they were restricted to run ning over a cell based ATM infrastructure, when the Internet was becoming increasingly packet–oriented.

Multilayer switching solutions.

Though multilayer solutions have many things in common, there are two fundamental different approaches to initiate the assignment and distribution of label binding to establish LSPs.

Data-driven model
Control driven model.

Data-driven Model:

First of all what is FLOW? A flow is a sequence of packets that have the same source and destination IP addresses and TCP or UDP port numbers. Now for label binding it all depends on the multilayer switch, as to when to create a label binding, it can be t he first packet or it can wait until it counts the number of packets. Merit of waiting for a number of packets is that you can figure out the overhead of assigning and distributing a label. MPLS does not support this Model.

Advantage
Label binding is created only if there is a flow that uses the label binding.

Disadvantage
When dealing with large ISP network, where there can be enormous number of individual traffic flow:

Many short-lived flows can create a burden on network operation, Control traffic needed to distribute label binding is directly proportional to number of traffic flows, Due to time difference in the recognition of a flow and assignment of a label to a flow, it should be taken care of so that packets that have not been assigned to a flow can be forwarded and not dropped, There should be some sophisticated packet classification capabilities to identify traffic flow.

Control-Driven Model:

In this model label bindings are created when control information arrives. Labels are assigned in response of either of these:

1) Normal processing of routing protocol traffic
2) Control traffic such as RSVP traffic
3) Static configuration

MPLS uses control-driven model.

Advantages of Control-driven Model are:

If a route exists in the IP forwarding table, it means a there is already a label for that route, so labels are assigned a immediately to the incoming traffic, Number of label switched paths is proportional to the number of entries in the IP forwarding table and not to the number of individual traffic flows. Hence scalability is better in this model, Label assignment and distribution overhead is lower in this model as label-switched paths are established only after a topology change or the arrival of c ontrol traffic and not with the arrival of each “new” traffic flow, Each packet is label switched and not like data driven model where only the tail end of the flow is label switched.

Problems with Multilayer Switching solutions:

The problems faced by multilayer switching solution was that it was proprietary as it maintained the IP control component and used ATM label swapping as the forwarding component. The major problem faced by this multi layer switching was that they couldn't operate over different architectures like LANs, PPP, Frame Relay etc., as they required an ATM transport. IETF designed MPLS to produce a unified and interoperable multi layer switching standards, as there should be some standard that could run over any link layer technology.

Solution

Multiprotocol label switching is built on the various multi layer-switching solutions. It is the latest step in Multilayer switching.

For building of LSPs, MPLS uses the control driven model to initiate assignment and distribution of the label bindings. In normal LSPs we consider traffic flow in one direction that is from head end towards the tail end. For duplex traffic we require t wo LSPs to carry traffic in each direction. LSPs can be built by merging one or more label switched hops which allows the packet to be forwarded from one label switch router (LSR-is a router that supports MPLS based forwarding) to another LSR across MPLS domain.

MPLS control component is similar to multi layer switching solutions. To support various platforms MPLS defines new label distribution protocols and standard – based IP signaling and some sort of extensions to existing protocols. MPLS does not sup port any ATM forum signaling or routing protocols. Hence coordinating two different protocol architectures is eliminated. MPLS is packet oriented.

MPLS forwarding component is based on label swapping algorithm. If in layer two there is a label field then this label field contains MPLS label. If layer two doesn't support a label field then MPLS label is in its standard MPLS header which is inserte d between IP header and layer two. MPLS allows any link layer technology to carry an MPLS label. So it can benefit from label swapping across an LSP.

The MPLS header is 32-bit and it contains the following fields:

1) Label field (20-bits): Contains the actual value of MPLS label.
2) CoS field (3-bits): It affects the queuing and discards algorithm applied to the packet as it is transmitted through the network.
3) The Stack (S) field (1-bit): It is used for hierarchial label stack.
4) The TTL (Time to Live) field (8-bits): It provides conventional IP TTL functionalities. Ref # 12

What will MPLS do?

The advantage of MPLS is that it provides a way to deliver new services that cannot be possible by conventional IP routing techniques without any increase in costs, it also provides better level of base service. MPLS provides enhanced routing capabilities e.g. if either host A or host B transmit a packet to host C. The packet follows path-1 across the core of the network. Because this is the shortest path computed.

Figure: MPLS enhances Routing Functionality Ref # 13.
 

Suppose there is congestion in router B, the policy would try to reduce this congestion by distributing the traffic load along different paths across the network. Traffic generated by A and destined to C would follow path-1. Various traffic generated a t host B and destined to C will follow part-2. Using conventional IP routing this cannot be implemented because all forwarding at router A is biased on the packet's destination address.

“To reduce congestion at LSR B, if the routers in the core of the network function as LSRs. The network administrator configures LSP 1 to follow Path 1 and then LSP 2 to follow Path 2. Finally, the network administrator configures LSR A to place a ll traffic received from Host A and destined for Host C into LSP 1. Similarly, LSR A is configured to place all traffic received from Host B and destined for Host C into LSP 2. The capability to assign any FEC to a custom-tailored LSP gives the network ad ministrator exact control of traffic as it is flowing through the provider's network.
 
Hence MPLS provides an unpredictable level of control over traffic, resulting in a network that is more efficient, supports more predictable services and can provide the flexibility required to meet constantly changing expectations. New services can be in corporated by simply changing the control component that assigns packets to an FEC and then maps each FEC to custom built LSP. ” Ref # 13.

MPLS application:

Presently there are three applications for MPLS in large ISP networks:

1) Traffic Engineering
2) Class of service (CoS)
3) Virtual private networks (VPNs)

Up to now, only circuit switched networks have provided the quality of service, performance guarantees, and reliability that many of today's mission critical applications require. But Multiprotocol label switching is changing all that by providing comp arable levels of service over IP networks.

Traffic Engineering:

In traffic Engineering we try to move the traffic flows away from the shortest path onto less congested physical path across the network.

This is currently the main objective of MPLS because of the unpredictable growth in demand for network sources. Successful traffic engineering can utilize all individual components like routers, switches and various links to their fullest extent, which results in a network that is more efficiently operated and provides more predictable services.

Applications:

ISPs understood that traffic engineering could significantly enhance the operation and performance of their network; hence they intended to use traffic engineering capabilities to:
To route paths around the points of congestion in network, How and which route should a traffic follow once the traffic is rerouted because of failures, Efficient use of bandwidth and ensuring that subsets of the network do not become either over utilized or underutilized, Maximizing operational efficiency, resulting in lowering operational cost, Minimizing packet loss, congestion and maximizing throughput, Provide more options, lower costs and better services to their customers.

Limitations of traditionally routed networks:

There are scalability issues with traditionally routed networks:
Traditional software routers could not handle heavy loads because their bandwidth and packet-processing capabilities were limited, Traditional solutions to traffic engineering problems were not scalable that is they offered a trial and error approach rath er than a scientific solutions to complex problems, Traffic loads were not taken into account while forwarding table was calculated, hence traffic was unevenly distributed across the network's links, causing inefficient use of resources.

MPLS helps in traffic engineering for the following reasons:

Supporting explicit paths allows network administrators to give the exact physical path that an LSP takes across the service provider's network, To identify bottlenecks and trunk utilization, and to plan for future expansion, Per-LSP statistics can be used as input to network planning and analysis tools, Constraint-based routing gives better capabilities that allow an LSP to satisfy certain requirements in performance before it is established.
 
An MPLS-based solution can run over packet-oriented networks and is not limited to ATM infrastructures.

Class of Service (CoS):

Some routers analyze the packets network layer header not merely to choose the packet's next hop, but also to determine a packets “precedence” or “class of service”. They may then apply different discard thresholds or scheduling dis ciplines to different packets. MPLS allows (but does not require) the precedence or class of service to be fully or partially inferred from the label. In this case one may say that the label represents the combination of FEC and a precedence or class of s ervice.

Voice and Video work best with circuit switching transport because it minimizes delay variations. MPLS will make the quality of voice and Video much easier to control by establishing proper per hop behavior for delay sensitive traffic. In addition, the reservation and dynamic placement options of Constraint-based Label Distribution Protocol (CR-LDP) will help keep this potentially disruptive traffic under control-unlike current UDP-based video, which can create problems if traffic is not controlled pro perly. For this ISPs need to also adopt traffic classification technology.

An ISP can take two approaches to support MPLS-based class of service forwarding:

1) According to the setting of the precedence bits carried in the MPLS header, the traffic flowing in a particular LSP can be queued for transmission on
      each LSR's outbound interface.

2) Between each pair of edge LSRs, An ISP can provide multiple LSPs between each pair of edge LSRs. To provide different performance and
      bandwidth guarantees, each LSP can be traffic engineered. The head end LSR places high-priority traffic in one LSP, medium-priority traffic in another
      LSP, best-effort traffic in a third LSP, and less-than-best-effort traffic in a fourth LSP.

Virtual Private Network (VPNs):

Virtual Private networks provide a mean of connecting remote users to host via the Internet thus eliminating the requirement of having a dedicated private network. By connecting through the Internet hardware and transmission costs are lowered.

Many customers outsource their Internet service related activities to Internet service Provider. Thus the customer need not worry about the hardware and software maintenance. There is drastic reduction in terms of cost while using long distance connect ion charges as compared to the rates offered by the ISP.

However this leads to lot of issues when allowing customers information and data being shared or accessed through the Internet. Information shared by customers should be with valid sites, which leads to authentication issues. Encryption is used to prov ide security for data while being transferred or traveling over the internet. Firewalls may be used to protect against security threats from outside world by a customer.

VPNs require controllable, efficient tunnels. MPLS transport does not look at the headers of the packets it is transporting therefore; the addressing used in those packets may be private. VPNs using MPLS provide many of the advantages like customers ca n choose there own addressing plans, which may or may not overlap with those of other customers or service providers. Each customer is confident that the data will be delivered anywhere but to sites within the customers VPN, Hence encryption is often not required. MPLS VPN model is highly scalable with increasing numbers of customers. MPLS provide simple and flexible tunneling mechanism. A VPN can be made by connecting a set of LSPs among different sites in the VPN. Each VPN site then tells the ISP about a set of reachable prefixes within the local site. VPN identifiers allow a single routing system to support multiple VPNs whose internal address spaces overlap with each other, Finally each ingress LSR places traffic into LSPs based on a combination of a packet's destination address and VPN membership information.

Goals:

The main aim of groups working on MPLS is to combine the use of label swapping in forwarding component with network layer routing in the control component. To achieve this, these groups have to satisfy a number of requirements like

It should run over any link layer technology and not just only ATM, The forwarding component of MPLS should support both unicast and multicast traffic flows, It should be compatible with all integrated service's models including RSVP, MPLS should be sc alable, MPLS must have features, which are present in current IP networks. It should support operations, administration and maintenance facilities present in IP networks.

Conclusion:

The various benefits such as predictability, scalability and manageability that service providers desperately need in their networks are provided by MPLS. Efforts are going on to extend MPLS to provide traffic engineering capabilities for service provider 's networks. This will allow the service providers to offer differentiated services. MPLS will require some modifications to existing equipment, but it will not require extensive equipment replacement.

The most important benefit of MPLS is that it allows ISPs to provide new services that cannot be supported by conventional IP routing techniques. MPLS gives better routing capabilities by supporting more than destination based forwarding. Traffic engin eering, CoS based forwarding, and VPNs are some of the new cost reduction and revenue generating services that can be deployed with MPLS. MPLS provides the flexibility to evolve control functionality without changing the forwarding mechanism, by separatin g the control component from the forwarding component. This uniquely positions MPLS to support enhanced forwarding capabilities that will be required for the Internet to grow.

The overall summary of MPLS is that it defines an evolutionary networking paradigm that combines the operating principles of layer 2 and layer 3 technologies while preserving service provider's investment in IP routing technology at the edge of the net work and switching technology in the core of the network.

Some believe that MPLS provide a standard that will allow transforming ATM switches into high performance Internet backbone routers. This was the original goal of multilayer switching solutions. But due to the recent advances in technology, there are c ertain look-up engines that run as fast as MPLS. Though MPLS can increase the forwarding performance of processor based system, but forwarding performance was not the primary goal of MPLS. Some believe MPLS was designed to eliminate the need of longest ma tch IP routing. Whereas this was not an objective of MPLS as layer-3 routing would always be required in the Internet.

One of the challenges of MPLS is service interruption during failure. Work is going on to solve this problem and various algorithms and protocols are there to minimize these LSP service interruption. Now work is going on to build some sort of hardware based local repair above two physically unrelated links. Thus hardware can easily switch from a bad to a good link when a failure occurs. Work is going on software side too that may repair at LSP level. This is a less expensive solution.

References: