Side_1_360

PATH-message is received by the destination

host, the RESV-message is created and transmit-

ted upstream towards the source host, making

the requested resource allocations in each router

along the path, if possible. The receipt of the

RESV-message in the source host signifies that

consistent resource allocations are being made in

every router, so that transmission of user data

may begin.

The main functional blocks of an IntServ router

are shown in Figure 4. When compared with the

Best Effort router, the forwarding module has

become more complex. It constitutes an MF

(Multi-Field) classifier, mapping the incoming

packets to a service class based on multiple

fields in the packet header. The scheduler is

managing the forwarding of flows from the out-

put queues of the Buffer Management block.

In addition to the modules for routing and for-

warding, there is a Signalling module dealing

with the RSVP signalling messages, and an

Admission Control module handling the re-

source allocation requests.

By basing the service provisioning on resource

allocation and policing (in the scheduler) of the

individual flows, IntServ is capable of providing

differentiated service classes in a “fair” manner

and with a quantifiable QoS. It overcomes the

first three limitations of the Best Effort service.

The main problem with IntServ is the extensive

processing taking place in each router due to the

per-flow handling of the resource allocation. For

the same reason, the establishment of new paths

to avoid bottlenecks in the network or support

load-sharing or protection switching is less

attractive, albeit possible.

2.3 Modelling the Network Topology

In the current topology model [GenIPmodel], the

router is represented on the network level by a

subnetwork with one connection point group

(CPG) constituting a number of equal connec-

tion points (CPs) for each incoming and outgo-

ing link1)as shown in Figure 5. Each CPG cor-

responds to a set of physical ports.

When compared to the best-effort router in

Figure 1, the function modelled is the routing.

Forwarding is not part of the topological repre-

sentation.

The interconnections between the routers are

modelled as links with a certain capacity in

terms of bandwidth. More than one interconnec-

tion may be transported over a single link.

The topological structure of an IP network may

be modelled on the basis of subnetworks and

links as shown in Figure 6. A Topological Link

is a link exhibiting the additional constraint of

being supported by a single trail in the server

layer. For example, the IP link in Figure 4 is

a topological link if it is supported by a single

LSP in the MPLS layer.

Subnetworks are either interconnected locally by

means of common CPGs or remotely via links.

2.3.1 Partitioning of Subnetworks

A more abstract representation of the network in

Figure 6 is obtained by hiding the internal struc-

ture represented by the contained links and sub-

networks, leaving the outer subnetwork and the

CPGs at the boundary of the outer subnetwork

only to be visible. This process may be recur-

sively repeated, creating a more abstract repre-

sentation each time. The inverse process, decom-

posing subnetworks into contained subnetwork

groupings is called subnetwork partitioning.

Figure 4 The IntServ router

Figure 5 Router, network view

1)In many cases, all the input ports belonging to one incoming link are interconnected so there is

only one input CP.

Routing

Module

Admission

Control

• •


• 
  

MF

Classifier Scheduler

Buffer Management Traffic Control

IP Forwarding module

Signalling

Module

CPG

(Input

Ports)

Subnetwork

(Router)

CPG

(Output

Ports)

CPG

(Input

Ports)

CPG

(Output

Ports)