Side_1_360

In case the IP-based network is placed over
another network, an overlay model is seen.
Then, the underlying network, e.g. WDM or
ATM, can be managed separately. However,
some initiatives are initiated to see the IP-layer
and an underlying network co-operating, like
IP over optics.

Traffic Engineering encompasses mechanisms
that control a network’s response to traffic
demands and events that affect the traffic carry-
ing capabilities. As mentioned earlier, TE
includes:

• Traffic management, e.g. by controlling rout-
  ing of traffic, used to maximise the network
  performance under varying traffic load pat-
  terns;


• Capacity management, e.g. by controlling
  resource configuration, used to design the net-
  work in order to minimise its cost while per-
  formance objectives are met.

Besides these, network planning can be said to

include planning and deploying node and trans-

port capacity in advance of the traffic changes.

Thus, these three activities can be said to interact

on three time scales.

Trade-offs Traffic management Routing table management Capacity management

TE methods applied TE methods considerably Control load comparable Design efficiency comparable

vs. not applied improving performance for the two cases for the two cases

Centralised vs. distributed Distributed control Control load comparable Design efficiency comparable

routing table control performance somewhat on per-node basis for for the two cases

better (more up-to-date the two cases

status information

Off-line/pre-planned (TDR) On-line control somewhat TDR and EDR control load SDR and EDR gives comparable

vs. online (SDR, EDR) better performance less than SDR design efficiency, both are better

routing table control than TDR

FR vs. TDR vs. SDR EDR/SDR performance FR/TDR/EDR have lower EDR/SDR design efficiency better

vs. EDR path selection better than TDR better control load than SDR than TDR better than FR

than FR

Multilink vs. two-link Multilink path selection Multilink path selection Multilink design efficiency better

path selection better under overload control load generally than two-link

Two-link path selection less than two-link path

better under failure. selection

Two-link path selection

lower call set-up delay

Sparse logic topology vs. Sparse topology better Sparse topology control Sparse topology design efficiency

meshed logical topology under overload. load generally less than somewhat better than multi-area

Meshed topology better meshed topology

under failure

Local status information vs. Local status performance Local status control load Design efficiency comparable for the

global status information somewhat better than global less than global status two cases

(more up-to-date information) control load

Status dissemination: Distributed query-for-status Centralized and distributed Design efficiency comparable for the

status flooding vs. somewhat better than status query-for-status two cases

distributed query-for-status flooding and centralized comparable on per-node

vs. centralised status status (more up-to-date basis. Status flooding

information) considerably higher control

load

Per-flow vs. per Comparable performance Per-virtual-network control Per-flow design efficiency somewhat

virtual-network (per-traffic load less than per-flow better than per-virtual network

trunk) traffic management control load

Integrated vs. separated Integrated network Total control load Integrated network design efficiency

voice and data network performance better than comparable for the better than separate network

separated network two cases

performance

Table 1 Trade-offs for

introducing TE-related

mechanisms (from the E.TE

series of recommendations)