MPLS.pdf - ethesis

This is to certify that the thesis entitled "A Framework for Optimized Bandwidth Allocation to LSPs in an MPLS Network" submitted by Sarah Panda : 107CS042 and Nitish Shukla : 107CS045 in partial fulfillment of the requirement for the degree of Bachelor of Technology in Computer Science Engineering, National Institute of Technology, Rourkela, is being conducted under my supervision. To the best of my knowledge, the matter contained in the thesis has not been submitted to any other university/institute for the award of any degree or diploma. We would also like to thank all the professors of the Department of Computer Science and Engineering, National Institute of Technology, Rourkela, for their guidance and inspiration.

Bandwidth allocation is a critical issue in the emerging MPLS technology in computer networking. To take into account the changing bandwidth needs and also the current data rate of the LSPs, this paper proposes a framework for fair bandwidth allocation to the LSPs in a more optimized manner. We compare the queue length parameter for all interfaces of all LSRs in the network and for a given interface at different data rate values.

Multiprotocol Label Switching (MPLS) is an emerging technology in the field of computer networking. It is a packet switching technology that has circuit switching features due to the introduction of virtual channels using labels.

Structure of an MPLS Cloud

Forwarding Equivalence Class(FEC)

Forwarding Table of an LSR/LER

MPLS Header

Working of a simple MPLS network

Weight of an LSP

In MPLS network, several LSPs can share the same link as shown in the figure below. LSP1 and LSP2 share the AB link, while LSP3 and LSP2 share the CE link. Link capacity must be fairly distributed among LSPs so that each of them can be used at any given time. A fair bandwidth allocation strategy in [1] and [2] was proposed earlier where the capacity is allocated in proportion to the weight carried by an LSP.

The above algorithm can be combined with the AIMD technique in MPLS as in [10] to add the benefits of both. One of the key aspects of MPLS is the addition of point-to-point path abstraction. This is done by the concept of label switched paths (LSP). Link abstraction supports constraint-based routing, which in turn is the basis for Generalized Multiprotocol Label Switching[5].

One of the applications of MPLS is constraint-based routing, which is often used to compute paths that satisfy certain requirements for a set of constraints. Some additional features have been added to GMPLS to manage some of the drawbacks in the MPLS control plane.

Figure 2.1: Multiple LSPs sharing a single link

Creating Label Switched Paths in MPLS network

Constraint Based Routing

Enhanced Link State IGPs

Compared to a normal IGP, an enhanced link state IGP floods information at more frequent intervals. Even without any change in topology, an improved link state floods IGP information due to change in link characteristics such as reservable bandwidth. So only when there is a significant change in bandwidth (above a certain predetermined threshold) does flooding occur.

Bandwidth allocation in MPLS networks

STATIC METHODS

DYNAMIC METHODS

Bandwidth allocation in MPLS NETWORKS 23 [3] proposed earlier is based on the Weighted Proportional Fair Rate Allocation Algorithm (WPFRAA) proposed in [2]. In this section, we summarize the operation of the WPFRA algorithm and then the one-way feedback-based algorithm. The quantity calculated here is the optimal bandwidth allocation for each LSP passing through the router in the forward direction, i.e.

When the control packet reaches the core router, the rf value is compared with the existing ER in the packet. The overhead involved in the round trip is significantly reduced in the one-way feedback based congestion control mechanism proposed in [3]. This algorithm [3] is based on the WPFRA mechanism given in [2] with the modification that the round trip of the control packet in the notification process is replaced by a one-way control loopback.

It is based on the fact that the path in an LSP in MPLS is reversible due to the use of labels. Since an inbound label is mapped to an outbound label through the information in the forwarding table, the reverse mapping is also possible due to the one-to-one correspondence between the labels. In this way, an LSP can be traced starting at the exit node in the opposite direction.

A 'down' LSP is an LSP that cannot handle traffic for some reason, but is defined in the forwarding tables. These types of algorithms fall under the category of OPEN LOOP problems, which include deciding when to accept new traffic, deciding when to discard packets and which ones, and making scheduling decisions at various points in the network. A PROPOSAL FOR EFFECTIVE BANDWIDTH ALLOCATION 29 The following table gives the values of some parameters for the four LSPs in the network.

The current data rate value is determined by observing the data traffic in the LSPs during the interval and determining the average data rate along the path. Bandwidth is allocated with ER values that depend on the LSP weight, which is an average value and may not always reflect the actual values. At intervals, a function called free bandwidth reset replaces any changes to LSP connections.

Figure 4.1: An MPLS network with four LSPs defined.

Changes to Control Packet

Free bandwidth is used to allocate more bandwidth to LSPs that wish to become active or whose bandwidth allocation is less than the requested bandwidth.

CASE : New LSP becomes active

CASE : An LSP gets deactivated

Once this is done, a free bandwidth reset is initiated for all LSPs whose egress router received the message. At the end of the interval, each LSP and node that had a potential change is updated with new free bandwidth.

CASE : Changing bandwidth allocation

FUNCTION : Reset free bandwidth()

Each router has two values for the free bandwidth of the LPi passing through it: - The actual free bandwidth of the link. The smallest of all free bandwidths of all connections up to that connection in the backward direction of the LSP. The message starts from that point and runs in the backward direction of the LSPs passing through it.

At each intermediate router, it now compares this new free bandwidth with the actual free bandwidth of the links and updates as necessary. In this way it reaches the entry node, with the new value of the free bandwidth. In the project properties bar we select the references tab and check the box 'inet' to indicate that the project will use the functions of the inet package.

Then we create a NED (network definition file) under that folder to define the network structure using the GUI tools in the framework. The NED file in the RSVPTE4.ned network example given below which we obtained from the inet/examples/mpls/testte tunnel folder for analysis looks like below. We create an 'rt' file for each host and LSR in the network, which represents the routing table information for the respective nodes.

For the network example, we have seven 'rt' files defined for LSRs and five for standard hosts. This is specifically part of an LSP from any of the hosts on the left end to host3. From this graph, it can be seen that as the data rate increases, the queue length first increases for small values of the data rate and then decreases steadily.

ANALYSIS a time increases and thus the traffic that accumulates at the initial node of the channel decreases. LSR5 is the only node through which any packet must pass to reach any of the destination hosts (host3 and host4). As we can see from the analysis given above, the queue length [1] of a node is highly dependent on the data rate of the channel through which it forwards the data.

In addition, the project addresses the issue of the optimized bandwidth allocation problem considering fluctuations in the bandwidth demand of LSPs. In this project, we have thus brought together the concept of varying LSPs' bandwidth replenishment and varying data rates to improve network bandwidth utilization.