Performance vs. Cost Analysis of WDM Networks with Dynamic Traffic Grooming

1
Performance vs. Cost Analysis of WDM Networks
with Dynamic Traffic Grooming Isabella Cerutti,
Andrea Fumagalli, Sonal Sheth Optical Networks
Advanced Research (OpNeAR) Lab The University of
Texas at Dallas
Traffic Grooming in WDM Networks
Performance vs. Cost Optimization Problems
Offline Network Design

• Given
• N number of nodes in the network
• L number of bidirectional lines in the
  network
• Hp average hop length of a lightpath
• find the optimal ratio of
• W number of wavelengths on each network
  line
• Ti /Ri number of transmitters/receivers at
  node i
• FG Ti Ri W of lines from/to node i
• SH Ti Ri
• MH Ti Ri , where H average hop
  length of a lightpath fixed a priori for
  different MH designs (1 H Hp )

• First Generation (FG) connections are
  multiplexed in any node (lightpaths are only
  between adjacent nodes)
• Single-Hop (SH) connections are multiplexed only
  at the terminal nodes of the end-to-end
  lightpaths
• Multi-Hop (MH) connections can be multiplexed at
  selected intermediate nodes

• Given a fixed budget to design a WDM network with
  grooming capabilities, which architecture
  provides the best performance?
• Offline problem Minimize the network resources
  to support a given traffic load
• Trade-off between reduced number of wavelengths
  (FG) and reduced number of transmitters/receivers
  (SH)
• Online problem Minimize the blocking
  probability of connection requests, for a given
  WDM network design

Lightpaths
SH
FG
MH
Online Connection Provisioning
Online Algorithm

• Based on auxiliary graph G(V,E)
• Add an edge between node pairs (i, j) to G(V,E)
  if
• A lightpath already exists and has bandwidth to
  accommodate the new connection request (type 1
  edge)
• A lightpath can be created, i.e., it is available
  a wavelength on any line along the shortest path
  between (i, j), a transmitter at node i and a
  receiver at node j (type 2 edge)
• Run shortest path on G(V,E) to route the
  connections, using resources cost as link weights

• Solve jointly the following sub-problems
• Virtual topology re-design
• Lightpath routing on the physical topology
  (shortest path algorithm)
• Wavelength assignment to the lightpath(s)
• End-to-end connection routing on the virtual
  topology

(wavelength ID, used capacity)
Physical topology
Example
Auxiliary graph G(V,E)
W 3, Ti Ri 6, connection bandwidth 0.2
Ring Network
Mesh Network
Conclusion

• For a fixed budget, MH outperforms SH and FG
• MH advantageous in poorly connected topology
  (ring) and with high wavelength grooming
  capabilities
• For a fixed budget, FG may outperform SH,
  depending on the line-to-node cost ratio and
  overall budget

?

• Connection blocking probability vs. normalized
  arrival rate ? (i.e., arrival rate/design cost)
• N L 14, Hp 3.77 , grooming factor g 10,
  line-to-node cost ratio ? 0.5

• Connection blocking prob. vs. ?, for ? 0.1 and
  0.9
• NSFNET topology N 14, L 21, g 10, Hp 2.14