Side_1_360

lem (see Box A). Furthermore, the reservation

information has to be conveyed between the

routers. The Resource reSerVation Protocol is

one means of doing this.

• Resource reSerVation Protocol (RSVP). RSVP
  is a signalling protocol allowing a receiver to
  initiate establishment of the reservation. It is a
  so-called soft state protocol in the sense that
  the reservation has to be refreshed repeatedly
  in order to keep the reservation for a longer
  interval. Both multicast and unicast flows are
  supported.


• Multi-Protocol Label Switching (MPLS).
  MPLS can be said to provide an additional
  forwarding mechanism. At the ingress of an
  MPLS domain, Label Switching Routers
  (LSRs) classify IP packets into Forwarding
  Equivalence Classes (FECs) based on certain

criteria. An MPLS label is then prepended to

the packets. Then, the subsequent LSRs may

only look at the MPLS label to decide upon

the forwarding treatment. A Label Switched

Path (LSP) is the path between an ingress LSR

and an egress LSR on which the packets are

sent. LSPs can be used for several purposes,

including load distribution, Virtual Private

Networks, and multicasting.

• Differentiated Services (DiffServ). Addressing
  the scalability challenge related to IntServ,
  DiffServ was proposed to categorise the traffic
  flows into a limited set of service classes. A
  class is defined using different classification,
  policing, shaping and scheduling policies.
  Hence aggregates of traffic flows are used,
  alleviating intermediate routers to consider
  individual traffic flows. A DiffServ field is
  defined in the IP header (part of the ToS octet,
  see [Jens01]) in order to indicate which ser-
  vice class a packet belongs to.


• Measurements. A set of metrics is developed
  by the IETF IP Performance Metrics (IPPM)
  working group. These can be used to monitor
  the performance observed by the end-users,
  network operators, and others. An architecture
  for handling measurements has been made by
  the IETF Real Time Flow Measurement
  (RTFM) working group. This architecture
  defines methods to do measurements of traffic
  flows and components involved. Such a sys-
  tem consists of meters, meter readers and
  managers.


• End-point congestion management. The IETF
  End-point Congestion Management working
  group is set to define congestion control
  mechanisms that transport protocols can use.
  A congestion manager monitors the paths of
  every congestion group under its control,
  using this information to distribute the capac-
  ity among the traffic flows in the group.
  Besides, procedures defined as part of TCP do
  also address congestion control, ref. [Jens01].

4 Requirements on TE

Systems

[ID_tepri] describes a number of requirements

that a TE system should fulfil. Here a require-

ment is understood as a capability needed to

solve a TE problem or to achieve a TE objective.

The requirements are either non-functional or

functional. A non-functional requirement relates

to the quality attributes of state characteristics of

a TE system. A functional requirement gives the

function a TE system should perform in order to

reach an objective or address a problem.

Box A Scalability

The term scalability is frequently observed in the discussion on various mecha-

nisms. Looking up in a dictionary, scalabiltiy is simply described as to allow

increase (or decrease) of something.

A common interpretation of scalability, however, is that the network equipment

may not be able to cope with the growing number of traffic flows or other units

used to express the amount. Hence, there is a threshold on the number of units

that a network node can handle. Such an understanding may implicitly assume

that the operating point is known (i.e. the range of number of units is given). Fur-

thermore, modifying the network node, e.g. by upgrading, would also affect the

capacity threshold.

A statement seen is that IntServ does not scale well, although this will likely fol-

low a linear growth as a function of the number of flows, see Figure Box A-1.

DiffServ, on the other hand, has an improved scalability as aggregates are

looked at. Introducing MPLS in combination with DiffServ may result in higher

demands on the nodes.

Figure Box A-1 Scalability concerns – for illustration only

Interpreting scalability may therefore ask for a certain tolerance in the sense that

inherent demands on the network nodes capacity are assumed.

max capacity
node N

Diffserv

max capacity
node K

Diffserv/MPLS

Intserv

Required node capacity

number of units,

e.g. traffic flows