Transcript Dynamic Traffic Engineering under Different Traffic Patterns

Multi-Layer Traffic
Engineering in IP over
Optical Networks
October 20, 2004
Hung-Ying Tyan
Department of Electrical Engineering
National Sun Yat-sen University
Outline
IP Network
 Transport Network
 Traffic Engineering

Some Observations
 Multi-Layer Traffic Engineering
 Our ML-TE Framework
 Our ML-TE Algorithms
 Evaluation

2
IP Network
Company
School
Enterprise
Internet Service Providers (ISP)
Carriers
Company
School
Enterprise
router
MAN, WAN
LAN
3
IP Network
(Optical) Transport Network
OXC
OXC
OXC
OXC
OXC
OXC
OXC
4
IP Network over OTN
Conceptual view
Data center
Actual
IP link = circuit
OXC
OXC
OXC
OXC
OXC
OXC
OXC
5
Transport Network

Evolved from traditional telecommunications networks

Good at long distance transmission of digital signal

Technologies




Synchronous Optical Network (SONET)
Wavelength Division Multiplexing (WDM)
Providing long-term circuits between end points
Separation from application networks


Network on network; “overlay network”
Business tiers: carriers vs ISPs
6
Traffic Engineering (TE)

Mechanisms to allocate network resources
according to traffic demand
 ISP:
Make better use of resources ($$$)

Static:
Network planning/provisioning/optimization

Dynamic:
Resources allocation adapts to traffic
change
7
Dynamic TE

Basic idea:
Move traffic around to alleviate congestion

Why is it effective?
 Data
traffic can be bursty
 Special events occur more frequently in data
networks
8
Observations

ISPs and carriers want to provide better
service with less cost

Over-provisioning because of slow
response to adding capacity and large
variation in traffic demand
 Utilization

< 25%
Current dynamic TE is still limited
 Only
deal with congestion
9
Large Daily Traffic Variation
OC-48 link between Dallas and Washington DC
10
More Observations

“Information Super Highway”?

Distribution channel of electronic
information products

Electronic post office
11
Technology Advances

Control Plane Technology
 A separate
network dedicated to resources control
 Allows resources to be added or released quickly

Optical devices and equipments
 Optical
laser, receiver, filter etc
 Wavelength conversion
 Optical add-drop multiplexer (OADM)
 Optical cross connect (OXC)
12
New TE Paradigm – Multi-Layer TE

For ISP, IP links can be leased or released on
demand
 IP
network topology can be changed on demand
 Let IP network topology adapt to actual traffic demand
2
3
2
4
2
4
Off-Peak Hours
3
Peak Hours
13
Value proposition

For ISP
 OPEX
reduction
 Simplified network planning

For Carrier
 New
applications/customers for Carrier
 Increased (overall) revenues
 Improved resource efficiency
 More revenue from the same resources
14
Network Model

Two-layer overlay
 IP/MPLS
network
 Optical network

OXC
OXC
Assume that Optical
TE is already
available
OXC
OXC
OXC
OXC
OXC
15
Our MLTE Framework
Input
-Traffic matrix
-Physical topology
-etc
Initial provisioning
Network monitoring
congestion
under-utilization
MPLS-TE
Hybrid path routing
no
Cost down?
yes
Remove
idle IP links
Activate
new IP links
16
Network Monitoring & MPLS-TE
IP/MPLS Network
1
2
3
1.
Monitor outgoing IP links


2.
3.
Detect congestion (if utilization > TH_high)
Detect underutilization ( if utilization < TH_low)
Select target LSPs and notify ingress nodes
Ingress node attempts to re-route LSPs
17
Hybrid Path Routing
Augmented
topology
information from optical
layer: candidate links
Hybrid
IP/MPLS Network
path consisting
of
 Existing
IP links
 Candidate links
OXC
OXC
OXC
OXC
Special
cost function
for both congestion
and under-utilization
OXC
OXC
OXC
Optical Network
Optical fiber
18

Hybrid Path Routing
Define
Network_Cost = sum( Link_Costi )
Link_Costi =
F(Link_Utilizationi) x real_link_costi
F

Algorithm:
 Triggered
by congested or underutilized links
UH link_utilization
 Dijkstra’s shortest path
 d(link_costi)= F(expected_link_utilizationi) –
F(link_utilizationi)
 Granting a new route only if it decreases the real
network cost
19
North America Model
OXC
Seattle
Chicago
OXC
Denver
SF
OXC
OXC
OXC
Cleveland
OXC
OXC
OXC
Detroit
OXC
OXC
Boston
NYC
DC
Kansas City
LA OXC
OXC
Dallas
Atlanta
OXC
OXC
14 Nodes
24 Links (fiber)
193 LSPs
Miami
LSP Demand
High
Low
0
8
16
Time (hr)
20
Experiment Results

Simulation tool: J-Sim (www.j-sim.org)

North America Model
 14
nodes, 24 links, 193 LSPs
 Average # of IP links ~ 21
 # of IP links at peak demand = 33
 Cost saving ~ 36% v.s. over-provisioning
 Tradeoff
between cost and number of LSP
reroutes
21
Visualization Tool
22
Research Topics

ML-TE framework

TE operations
 MPLS-TE
procedure and Optical-TE

Topology transformation algorithm

Hybrid path routing algorithm
 Suitable
for both congestion and under-utilization
23
Thank you!
Question?
24
Start
Initial network provisioning:
Set up LSP’s for initial demand
Network monitoring
no
Is any IP link congested
or under-utilized?
Basic workflow
yes
LSP selection on the target IP link
Hybrid path routing computation
for re-routing the selected LSP
no
Network cost reduced if
this re-route is performed?
yes
Does the hybrid
path contain
“candidate links”?
yes
Activate new IP
links on those
candidate links
no
LSP rerouting
Does the re-route
result in idle IP links
in the network?
no
yes
Remove idle IP
links
25