Traffic Grooming - Rudra Dutta, Associate Professor

Download Report

Transcript Traffic Grooming - Rudra Dutta, Associate Professor

Traffic Grooming
Apoorv Nayak
Prathyusha Dasari
Agenda

Improved approaches for cost effective traffic
grooming in WDM ring networks





Motivation
Terminology
Single hop approach
Multi hop approach
A Novel Generic Graph Model for Traffic
Grooming in Heterogeneous WDM Mesh
Networks
Motivation

With WDM technology we can have dozens of
wavelengths on a fiber.

Increase in network capacity is accompanied
with increase in the electronic multiplexing
equipment.

Dominant cost is electronics and not fiber.
Aim

Goal is to minimize electronic costs by reducing
the number of ADM’s and make efficient use of
wavelengths.

“Groom” a number of low rate traffic streams
onto a higher rate stream and vice versa.

Reducing the number of wavelengths
Terminology

SONET

ADM

WADM
SONET Ring

Much of today’s physical layer infrastructure is
built around SONET rings.

Constructed using fiber (one or two pairs usually
used to provide protection) to connect SONET
ADM’s.
Example
Signal from A split into two; one copy transmitted over the working ring
(1) other copy over protection ring (8-7-6).
B selects the best signal.
SONET ADM

Add/Drop multiplexer.

Each ADM can multiplex multiple lower rate
streams to form a higher rate stream OR
demultiplex a higher rate stream to several
lower rate ones.

Employs O-E-O conversion.

Works at a particular wavelength.
Example
M<N
WADM

Wavelength add/drop multiplexer.
 Emergence of WDM technology has enabled a single
fiber pair to support multiple wavelengths.
 Since ADM works on a single wavelength, if there are W
wavelengths, every node would need N*W ADM’s.
WADM contd
But a node may not need to add / drop streams on
every wavelength.
WADM’s can add/drop only the wavelengths carrying
traffic to/ from a node.
Example of a SONET ring
OC-48 SONET ring
Assumptions

Traffic demands are static and known a priori.

Traffic is uniform;total bandwidth required is
same for any s-d pair.

Unidirectional ring considered.
Single hop approach

Uses the simulated annealing heuristic.

A node with a wavelength-k ADM can communicate
directly with all other nodes having wavelength-k ADM.

Formation of a wavelength-k logical ring which consists
of the subset of N nodes with a wavelength-k ADM.

Nodes within a logical ring communicate with each other
directly (single hop).
Logical Rings
1
1
5
3
2
3
4
1
2
3
2
Example of single hop approach
Given data





Network layout
Traffic demand matrix
Number of available wavelengths : 2
Capacity of each wavelength : OC-3
Uniform traffic between any two nodes is OC-1.
Network Topology
a) Physical Network
0
1
b) Traffic on the Network
t1
0
t5
1
t6
fiber
t2
t4
3
2
3
t3
2
Traffic Grooming Approach1 (Random)
Total number of ADM’s needed = 8
Traffic Grooming Approach 2
Total number of ADM’s needed = 7
Single hop traffic grooming algorithm
do{
do{
dcost = perturb();
if(∆cost < 0 or (∆cost > 0 and exp(-∆cost/control) > rand [0,1)))
{
accept_change();
chain++;
}
else
reject_change();
} while(chain < ANN_CONST * G)
control = control * DEC_CONST;
} while(control > END)
Terminology
Perturb() – Randomly swap positions of two circles in different
wavelengths.
ANN_CONST– Decides how long to run the algorithm before
system reaches equilibrium.
DEC_CONST - How fast to lower the control variable.
G – Grooming ratio (Ratio of the wavelength channel rate to the
lowest traffic rate).
Multi hop approach (Hub based communication)
 Source and destination on different logical rings.
Solution
 OXC?
Still maturing
Costly
 Multiple ADM’s?
 Relatively inexpensive as compared to OXC
 More delay and reduced throughput
 Price-Performance tradeoff.
Approach followed in paper?
 A “hub” node with an ADM for each wavelength.
 Multiple ADM’s at some nodes.
 Decide which nodes, how many ADM’s, which
wavelengths.
Example of a unidirectional multihop WDM ring network
Terminology and assumptions
 W - Number of wavelengths.
 Di - Number of ADM’s in the ith node.
 G - Grooming ratio (Ratio of the wavelength channel
rate to the lowest traffic rate).
 tij - Traffic requirement (Number of low rate circuits
between i and j for i-j pair).
 tij = 1 for uniform traffic.
Given data
Number of nodes- N
Traffic matrix- T
Grooming ratio- G
ADM placement algorithm
Input N, G, t;
Compute number of ADM’s needed at each node by the equation:
Compute number of wavelengths by the equation:
Create an ADM hub node;
Place ADM’s needed at each node sequentially;
While (no of ADM’s and wavelengths can be reduced)
{
Assign traffic on each wavelength using shortest path;
Traffic grooming (wavelength combining and segment swapping);
}
Wavelength Combining
If capacity (i) < G and capacity (j) <G
And capacity (i) + capacity (j) ≤ G
Then the two wavelengths can be combined
W1 =2
1
W1 =1
W2 =2
W2 =1
W1=3
W1=3
4
2
W1 =2
W2 =1
W1=3
W1 =1
W1=3
W2 =2
3
Segment Swapping
Helps in wavelength combining by “manipulating” wavelengths such
that all link capacities are less than G.
1
W1
W1=2
=3
W2
=3
W1=3
W2
=2
W1 = 2
W2 = 3
W1=3
W3=1
W2=3
W3=1
W2=3
W3=1
4
2
W1
=2=3
W1
W1=3
W2
=3
W2
=2
W2=3
W3=1
W3=1
W1 =2
W2 =3
W1=3
3
W3=1
W2=3
Example of multihop approach
Given data
N=5, node 0 is hub node
G = 3, tij = 1
By eqn 1, every node needs at least 2 ADMs
By eqn 2, total number of wavelengths is 8
0
1
2
3
4
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
1
1
1
1
3
1
3
1
3
1
3
1
2
1
2
1
2
1
2
1
1
1
1
1
Final result
Number of wavelengths = 4
Number of ADM’s = 12
Comparisons
Increase in G, decrease in W, less ADM’s in hub node
A Novel generic Approach






Objectives
Generic Graph Model
Auxiliary Graph
- Vertices
- Edges
IGABAG
Example
Grooming Policies
- Comparison
In Heterogenenous Networks
Traffic Grooming :
• Can be applied to static or dynamic traffic grooming problem.
• Each node is characterized by various parameters
- Optical switching/multiplexing capabilitieswavelength/waveband/fiber.
- Electronic switching/multiplexing grooming
capabilities.
- Availability of wavelength conversion.
- Number of transmitters/receivers.
CSC 778 Fall 2007
Objectives
• Traffic grooming problem may have various objectives
- Minimize cost (transmitters/receivers).
- Minimize overall traffic load.
- Minimize maximum traffic on any light path.
- Minimize maximum wavelengths on any fiber.
CSC 778 Fall 2007
Generic Graph Model
• Construct auxiliary graph
 Add vertices and edges corresponding to network
elements.
- Links
- Wavelength converters
- Electronic ports (transmitters/receivers)
 Assign costs to links based on objective
• Run shortest path algorithm
CSC 778 Fall 2007
Auxiliary Graph - Vertices
• Input and output vertex for each wavelength layer at each node
• Input and output vertex for lightpath layer at each node
• Input and output vertex for access layer at each node
CSC 778 Fall 2007
Auxiliary Graph - Edges
• Wavelength Bypass Edges (WBE)
- From each input to output port on a given
wavelength layer.
- Optical wavelength switching capability
• Grooming Edges (GmE)
- From input to output port on access layer if
grooming is available.
- Electronic switching capability.
• Mux Edges (MuxE)
- From output port on access layer to output port
on lightpath layer.
• Demux Edges (DmxE)
- From input port on lightpath layer to output port on access layer.
CSC 778 Fall 2007
Auxiliary Graph - Edges
• Transmitter Edges (TxE)
- From output port on access layer to output port on
wavelength layer if transmitter is available .
• Receiver Edges (RxE)
- From input port on wavelength layer to input port on
access layer if receiver is available.
• Converter Edges (CvtE)
- From input port on wavelength layer  1 to output
port on wavelength layer  2 if optical wavelength
conversion is available.
CSC 778 Fall 2007
Auxiliary Graph - Edges
• Wavelength-Link Edges (WLE)
- From output port on wavelength layer l at node i to
input port on wavelength layer l at node j if
wavelength l is available on the physical link between i
and j
• Lightpath Edges (LPE)
- From output port on the lightpath layer at node i to the
input port of the lightpath layer at node j if there is a
lightpath from node i to node j
CSC 778 Fall 2007
.
Auxiliary Graph - Edges
GrmE
DmxE
MuxE
TxE
CvtE
RxE
WBE
WLE
CSC 778 Fall 2007
Integrated Grooming Based on the Auxiliary Graph
(IGABAG)
• Traffic demand: T(s,d,g,m)
- s : source, d : destination, g: granularity,
m: amount of traffic in units of g
• Step 1: Delete edges with capacity less than g.
• Step 2: Find shortest path p from output port on the access
layer of s to the input port on the access layer of d.
• Step 3: If p contains wavelength-link edges, set up
corresponding lightpaths.
• Step 4: Route traffic demand along path p. If the capacity
of lightpaths along p is less than m, route the maximum
amount possible.
• Step 5: Restore edges deleted in Step 1.
• Step 6: Update graph G.
CSC 778 Fall 2007
Example
• Wavelength capacity: OC-48
• Each node has 2 transmitters/receivers
• Granularity: OC-12
• Request 1: T(1, 0, OC-12, 2)
-> Lightpath on  1 from N1
to N0
• Request 2: T(2, 0, OC-12, 1)
• Request 3: T(1, 0, OC-48, 1)
CSC 778 Fall 2007
Example..
CSC 778 Fall 2007
Example…
0
1
2
CSC 778 Fall 2007
Example – Single-Hop Grooming
• Request 2: T(2, 0, OC-12, 1)
- new lightpath on  2 from N2-N1-N0
• Request 3: T(1, 0, OC-48, 1)
CSC 778 Fall 2007
Example: single-hop grooming
CSC 778 Fall 2007
Example: single-hop grooming
0
1
2
CSC 778 Fall 2007
Example – Multi-hop Grooming
• Request 2: T(2, 0, OC-12, 1)
- new lightpath on  1 from N2-N1
- Existing lightpath on  1 from N1-N0
• Request 3: T(1, 0, OC-48, 1)
CSC 778 Fall 2007
Example: multi-hop grooming
CSC 778 Fall 2007
Example: multi-hop grooming
0
1
2
CSC 778 Fall 2007
Grooming Operations
Add New
Lightpath(s)
Single-hop or
multi-hop
grooming
Operation 1
No
Single-hop
Operation 2
No
Multi-hop
Operation 3
Yes
Single-hop
Operation 4
Yes
Multi-hop
CSC 778 Fall 2007
Grooming Policies
• Minimize the Number of Traffic Hops (MinTH)
- Attempt Operation 1
- Attempt Operation 3
- Between Operation 2 and 4, choose the one with fewest logical
hops
• Minimize the Number of Lightpaths (MinLP)
- Attempt Operation 1
- Attempt Operation 2
- Attempt Operation 3 or 4
• Minimize the Number of Wavelength-Links (MinWL)
- Attempt Operation 1
- Attempt Operation 2
- Between Operation 3 and 4, choose the one with fewer wavelength
links
CSC 778 Fall 2007
Ordering of Requests for Static Case
• Least Cost First (LCF)
- Establish least-cost request first
- Cost = (weight of shortest path for demand)/(amount of traffic)
• Maximum Utilization First (MUF)
- Select connection with maximum utilization first
- Utilization = (amount of traffic)/(number of hops on physical
topology)
• Maximum Amount First (MAF)
- Select connection with largest traffic demand first
CSC 778 Fall 2007
Comparison of Policies – Non Blocking Model
CSC 778 Fall 2007
Comparison of Policies – Blocking Model
CSC 778 Fall 2007
Acknowledgments
• Improved Approaches for Cost-effective traffic grooming in WDM Ring Networks
: Uniform-Traffic Case; Wonghong Cho, Jian Wang, Biswanath Mukherjee
• A Novel generic graph model for traffic grooming in heterogenous WDM mesh
networks; Hongyue Zhu, Hui Zang, Biswanath Mukherjee
•Traffic grooming algorithms for reducing electronic multiplexing costs in WDM
rings; Angela L. Chiu, Eytan H. Modiano
• An effective and comprehensive approach for traffic grooming and wavelength in
SONET/WDM rings; Xijun Zhang, Chunming Qiao
•Improved approaches for cost-effective traffic grooming in WDM ring networks:
Non-uniform traffic and bidirectional ring; Jian Wang, V. Rao Vemuri
•Connection Oriented Networks, Harry Perros
Questions??