Research Issues on Routing and Research Issues on Routing and Wavelength Assignment for Wavelength Wavelength Assignment for Wavelength Routed WDM Networks Routed WDM Networks Hsu-Chen, Cheng PhD. Student Department of Information Management National Taiwan University 10/27/2003 Qualifying Exam Qualifying Exam
62
Embed
Research Issues on Routing and Wavelength Assignment for Wavelength Routed WDM Networks Hsu-Chen, Cheng PhD. Student Department of Information Management.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Research Issues on Routing and Wavelength Research Issues on Routing and Wavelength Assignment for Wavelength Routed WDM NetworksAssignment for Wavelength Routed WDM Networks
Hsu-Chen, Cheng
PhD. Student
Department of Information Management
National Taiwan University
10/27/2003
Qualifying ExamQualifying Exam
2
OutlineOutline
Introduction of Optical Networks WDM Technology Optical Network Control Plane IP/WDM Traffic Engineering
Optical Network Design and Engineering Routing and Wavelength Assignment (RWA) Heuristics
Optical Multicasting Multi-granularity Architecture of Optical Network Future Research Direction
3
The Trend of Network TechnologyThe Trend of Network Technology
Circuit Switch Packet Switch
E
M
S
DWDM
SONET
ATM
IP
E
M
S DWDM
SONET
IP /MPLS E
M
S DWDM OXC
Thin SONET
IP GMPLS
E
M
S
DWDM OXC
IP GMPLS
Time
Introduction of Optical Networks
4
Optical NetworksOptical Networks
First Generation: FDDI Gigabit Ethernet
Second Generation: WDM Local Area Network
Passive Star Network Single-Hop WDM Network
WDM Wide Area Network Wavelength routed Network OBS OPS
Wavelength-routed optical networks are the most promising candidates for backbone high-speed WAN.
Overlay Model Closer to classical IP and ARP over ATM scheme The IP and optical network are independent of each other Edge IP router interacts with its ingress OXC over a well-
defined UNI Peer Model
The IP and optical network are treated together as a single network
Augmented Model IP and optical networks use separate routing protocol, but
information from one routing protocol is passed through the other routing protocol
Introduction of Optical Networks
G. N. Rouskas and H. G. Perros, A Tutorial on Optical Networks, Networking 2002 Tutorials, vol. 2497, 2002, pp. 155-193.
12
TE ModelTE Model
WDM Traffic Engineering Model Minimum average packet delay Maximize scale up capability
MPLS Traffic Engineering Model Overlay model Virtual topology (LSPs)
IP over WDM Traffic Engineering Model Virtual topology design Routing and wavelength assignment
This step includes – Sizing the links (no. of wavelength channel, capability of eac
h channel) Sizing the OXCs Placement of resources (Amplifiers, Converters, Splitters) Dealing with link or OXC failures
Placement of converters [J. Iness, 1999] Sparse location of wavelength converters Sharing of converters (Nodal Design) Limited-range wavelength conversion [R. Ramaswami and G.
H. Sasaki, 1998]
Static RWA
17
Virtual Topology DesignVirtual Topology Design
A solution to the static RWA problem consists of a set of long-lived ligthpaths which create a logical topology among the edges node.
It is not possible to implement fully connected virtual topologies.
N(N-1) lightpaths Objective
Minimize the maximum congestion level Minimize the average weighted number of hops Minimize the average packet delay
Static RWA
18
Wavelength AssignmentWavelength Assignment
1
23
4
5
6
7
8
Static RWA
1
3
4
56
7
8
2
1λ
1λ
1λ
0λ
0λ
0λ
2λ 2λ
Graph-coloring problem
NP-complete
Sequential graph-coloring algorithms
19
Route ComputationRoute Computation
Static algorithm and adaptive algorithm Lightpath routing
Constraint algorithm Fixed routing algorithm Fixed alternative algorithm
General Model of Virtual Topology DesignGeneral Model of Virtual Topology Design
Mixed Integer Linear Programming [B. Mukherjee et al., 1994] [B. Mukherjee et al., 1996] [R.Ramaswami and K. N. Sivarajan, 1996] [R. Krishnaswamy and K. N. Sivarajan, 1998] [R. Krishnaswamy and K. N. Sivarajan, 2001]
Objective
General Model
)min( max
23
CongestionCongestion
Congestion may be viewed as a function of the various parameters of the network such as
the traffic matrix, number of wavelengths the fiber can support, resources at each node (number of transmitters and
receivers), the hop lengths of the logical links, the multiplicity restrictions on the logical topology, the multiplicity restrictions on the physical topology, symmetry/asymmetry restrictions, the propagation delay.
General Model
24
ObjectiveObjective
The motivation for choosing this objective is that the electronic processing (switching speed) requirement is proportional to the congestion.
If the switching speeds at the nodes are limited, then minimizing congestion would be appropriate as it would enable the traffic carried per wavelength to increase.
If there is heavy traffic between some source–destination pair, then there is a logical link between them; this is a desirable property.
This happens because of the objective function, i.e., if there is heavy traffic between node i and node j then because of the objective there would tend to be an edge in the logical topology.
)min( max
General Model
25
NotationsNotations
s,d source and destination of a packet, when used as superscripts;
i,j originating and terminating node of a logical link (lightpath);
The above constraint ensures that the number of logical links originating (out-degree) and terminating (in-degree) at node is less than or equal to the number of transmitters and receivers at that node.
Let logical link bij use wavelength k. Then by conservation of wavelength constraints there is a path in the physical topology from node i to node j with wavelength assigned to it.
kmlmlcij
kij ),,( ,1),()(
mj)(i
jmim
im
jm
b
b
PlmCPmlC ij
ijF
k llm
kij
W
k llm
kij ,,
. and if
, if
if
,0
,
,
),(),(1
0,
)(1
0
)(
General Model
32
Constraints- Hop Bound ConstraintsConstraints- Hop Bound Constraints
[Y. Sun et al., 2001] The USCH1 algorithm gives the worst network throughput
The wavelength continuity constraint limits the performance of the MSCH1 algorithm
The best approach is the MSCH2 if the wavelength converters are not available
Optical Multicasting
47
Super-Lightpath RWASuper-Lightpath RWA
WDM + OTDM
Optical Multicasting
48
Tree Shared MulticastTree Shared Multicast
[D. N. Yang and W. Liao, 2003] A light tree can carry data of multiple multicast streams, and
data of a multicast stream may traverse multiple light-trees to reach a receiver.
Multicast routing and wavelength assignment of light-trees Design of light-tree based logical topology for multicast strea
ms [郭至鈞,民國 92]
Multicasting group aggregation and MC-RWA Source-based tree aggregation Maximize the total revenue Lagrangian relaxation
Optical Multicasting
49
Tree-Sharing StrategiesTree-Sharing Strategies
Given the set Mi at edge router i, we consider a strategy to decompose Mi into a number of MSCs (Multicast Sharing Class)
Perfect Overlap (PO), Super Overlap (SO), and Arbitrary Overlap (AO)
Optical Multicasting
50
Optical Waveband SwitchOptical Waveband Switch
WBS has attracted attention for its practical importance in reducing port count, associates control complexity, and cost of photonic cross-connect.
In WBS networks, several wavelengths are grouped together as a band and switch as single entity using single port.
MG-OXC not only switch traffic at multiple granularities such as fiber, band, and wavelength, but also add and drop traffic at multiple granularities .
End-to-end grouping: Grouping the traffic (lightpaths) with same source-destination
only One-end grouping:
Grouping the traffic between the same source (or destination) nodes and different destination (or source) nodes
Subpath grouping: Grouping traffic with common subpath (from any source to a
ny destination)
Multi-granularity Optical Networks
54
WRN vs. WBSWRN vs. WBS
WRN Minimum the number of wavelengths Minimum wavelength hops
WBS Minimum the number of ports
Waveband conversion
Multi-granularity Optical Networks
1λ
2λ
3λ
0λ
1λ
2λ
3λ
0λ
1λ
2λ
3λ
0λ
1λ
2λ
3λ
0λ
0b
1b
0b
1b
55
WBS Failure RecoveryWBS Failure Recovery
Band Merging
Band Swapping
Multi-granularity Optical Networks
1λ
2λ
3λ
5λ4λ
0λ
0b
1b
1λ
2λ
3λ
5λ4λ
0λ
0b
1b
56
Hierarchical Routing ModelHierarchical Routing Model
Network node architecture Sequence of routing and waveband aggregation Route Computation
Multi-granularity Optical Networks
Integrated routing
Separate routing
Offlinerouting
Onlinerouting
57
Researches of M. Lee et al. Researches of M. Lee et al.
Multi-Layer MG-OXC The waveband is formed by grouping lights with the same dest
ination in a network ILP formulation Maximize the reduction gain of crossconnect size with the mini
mum number of wavelengths Results
The introduction of waveband leads to a very large reduction in crossconnect requirements for large-scale networks.
A large reduction of crossconnect requirements can still be expected even at nonoptimal wavelength granularity.
The reduction depends on network topology, traffic demand and traffic pattern.
Multi-granularity Optical Networks
Integrated routing
Separate routing
Homog
eneo
us
netw
ork
Heter
ogen
eous
netw
ork
Offlinerouting
Onlinerouting
58
Researches of Y.Suemura et al. Researches of Y.Suemura et al.
Propose and analyze two heuristic routing and
aggregation algorithms (online and offline) to be used for homogeneous networks in separate routing framework.
Minimum the routing cost Assume that all the ports (OEO and optical ones) have the same
cost. The cost of routing is the total number of used ports.
The simulations demonstrate a significant cost reduction by employing hierarchical routing (from 33% in online algorithm to almost 60% in offline one)
Multi-granularity Optical Networks
Integrated routing
Separate routing
Homog
eneo
us
netw
ork
Heter
ogen
eous
netw
ork
Offlinerouting
Onlinerouting
59
Researches of X. Cao et al. Researches of X. Cao et al.
This research show that WBS is different from traditional wavelength, and thus techniques developed for wavelength-routed networks cannot be directly applied to effectively WBS-related problem.
The objective is to route lightpaths and assign appropriate wavelength to them so as the minimum the total number of prots required by the MG-OXCs.
Static offline problem (Network Planning) Balanced Path Routing with Heavy Traffic (BPHT)
Dynamic real-time problem (Network Servicing) Maximum Overlap Ratio (MOR)
Results BPHT: 50% fewer total ports than using ordinary OXCs MOR: 35 % saving in the number of ports
Multi-granularity Optical Networks
Integrated routing
Separate routing
Homog
eneo
us
netw
ork
Heter
ogen
eous
netw
ork
Offlinerouting
Onlinerouting
60
Researches of P. Ho et al. Researches of P. Ho et al.
Dynamic tunnel allocation (DTA) Use fixed alternative routing with k-shortest paths to inspect netwo
rks along each alternative path for dynamically setting up lightpaths.
Capacity-balanced static tunnel allocation (CB-STA) Fiber and waveband tunnels are allocated into networks at the pla
nning stage according to weighted network link-state.
Simulation Results DTA is outperformed by CB-STA in the same network environment
duo to a well-disciplined approach for allocating tunnels with CB-STA.
The mix of the two approaches yields the best performance given the same network environment apparatus.
Integrated routing
Separate routing
Homog
eneo
us
netw
ork
Heter
ogen
eous
netw
ork
Offlinerouting
Onlinerouting
Multi-granularity Optical Networks
61
Future Research TopicsFuture Research Topics
New optical component application Optical Multicasting
QoS Multicasting Tree Aggregation Problem Call Admission Control