Title: High-Availability%20Network%20Architectures%20(HAVANA):%20Comparative%20Study%20of%20Fully%20Pre-Cross-Connected%20Protection%20Architectures%20for%20Transparent%20Optical%20Networks%20Contact:%20grover@trlabs.ca
1High-Availability Network Architectures
(HAVANA)Comparative Study of Fully
Pre-Cross-Connected Protection Architectures for
Transparent Optical NetworksContact
grover_at_trlabs.ca
A. Grue, W. D. Grover, J. Doucette, B. Forst, D.
Onguetou, D. Baloukov TRLabs (Network Systems
Group) 7th Floor, 9107 116 Street Edmonton,
Alberta, Canada T6G 2V4
M. Clouqueur, D. Schupke Nokia Siemens
Networks (Network Control-Plane and
Transport) Otto-Hahn-Ring 6 81730 Munich, Germany
2Pre-Cross-Connection A Design Constraint
- Non-Pre-Cross-Connected
- Shared pool of spare capacity
- Backup paths cross-connected at failure time
- Examples SBPP, span-restorable mesh
- Pre-Cross-Connected
- Cross-connections for backup paths formed in
advance of failure - Resulting chains of pre-cross-connected capacity
coalesce into protection structures - Examples BLSR, p-cycles
x2
1
2
5
8
11
5
9
12
7
6
4
10
3
1
4
x5
x3
3Outline
- Architectures
- Project Overview
- Methods and Results
- Conclusions
4p-Cycles
Straddling span
On-cycle span
5Failure Independent Path Protecting p-Cycles
Straddling path
On-cycle path
6PXTs (Pre-Cross-Connected Trails)
Understanding PXTs Behave like FIPP cycles,
only the structures are not closed
As a consequence, they are not able to provide
two protection paths for failed working paths
7DSP (Demand-Wise Shared Protection)
Understanding DSP It is essentially 1N APS over
N1 disjoint routes between end nodes
8Outline
- Architectures
- Project Overview
- Methods and Results
- Conclusions
9Project HAVANA Outline/Objectives
- Objective To characterize and compare many
different pre-cross-connected protection
architectures on a single network, under
real-world constraints to network intelligence
and flexibility - Project Phases
- Basic architecture design (capacity for single
span failure restorability) - Dual failure analysis of basic designs
- Wavelength assignment feasibility and methods
- Optical path length constraints analysis and
enhancement - Outputs
- A set of best feasible network designs
- Theoretical insights into architectural
properties - Design methods and insights
10Outline
- Architectures
- Project Overview
- Methods and Results
- Basic architecture design
- Conclusions
11TestSet0 Network
12Working Routing Constraints
- Models for FIPP, PXTs, and p-cycles are SCP
(spare capacity placement) only working routing
is static - Both FIPP and PXTs require a working routing such
that at least one path, disjoint from the working
path, exists between the end nodes
13Results Spare Capacity Redundancy
- p-Cycles are the most capacity efficient
- DSP has capacity efficiencies just slightly lower
than that of 11 APS
14Outline
- Architectures
- Project Overview
- Methods and Results
- Basic architecture design
- Dual failures
- Conclusions
15Dual Failures Network Intelligence
- The response to a first failure cannot change as
a result of a second failure failure responses
are independent
16Results Dual Failures
100
of all failed paths restored over all dual
failure scenarios
DSP 85
PXTs and p-cycles 66
FIPP p-cycles 50
17Outline
- Architectures
- Project Overview
- Methods and Results
- Basic architecture design
- Dual failures
- Wavelength assignment
- Conclusions
18Wavelength Assignment in p-Cycles
- p-Cycles require either wavelength conversion or
at least 2 fibres on every span in order to
support wavelength continuity
Different wavelengths for 2 different working
paths
Wavelength conversion required for break-in
19Results Wavelength Assignment
- Wavelengths are allocated to the network in bands
of 20 - 40-wavelength (2 bands) assignment found for all
architectures - 20-wavelength (1 band) assignments found for
- PXTs (modified SCP model)
- FIPP p-cycles (JCP model necessary)
- Not found for
- DSP (impossible)
- p-cycles (perhaps possible using JCP?)
20Outline
- Architectures
- Project Overview
- Methods and Results
- Basic architecture design
- Dual failures
- Wavelength assignment
- Optical path lengths
- Conclusions
21Results Optical Path Lengths
- Only DSP design satisfied reach constraints with
the original design - PXTs and FIPP p-cycle designs easily found by
modifying the pre-processing step - Compliant p-cycle design found by using a new ILP
model altogether
22Outline
- Architectures
- Project Overview
- Methods and Results
- Conclusions
23Conclusions
Dual Failure Restorability
Wavelength Assignment
Optical Reach
Cost of Design
PXTs
p-Cycles
DSP
DSP
Best
FIPP p-Cycles, PXTs
FIPP p-Cycles, PXTs
PXTs, p-Cycles
FIPP p-Cycles
DSP, p-Cycles
p-Cycles
DSP
Worst
FIPP p-Cycles
24To Find Out More
- References on PXTs, FIPP p-Cycles, DSP (listed in
paper) - A. Kodian, W.D. Grover, Failure Independent
Path-Protecting p-Cycles Efficient and Simple
Fully Pre-connected Optical-path Protection,
IEEE Journal of Lightwave Technology, vol. 23,
no.10, October 2005. - T. Y. Chow, F. Chudak, A. M. Ffrench. Fast
Optical Layer Mesh Protection Using
Pre-Cross-Connected Trails, IEEE/ACM Trans.
Networking, vol. 12, no. 3, pp. 539-547, June
2004. - Koster, A. Zymolka, M. Jager, R. Hulsermann,
Demand-wise Shared Protection for Meshed Optical
Networks, Journal of Network and Systems
Management, vol. 13, no. 1, pp. 35-55, March
2005. - A. Grue, W.D. Grover, Characterization of
pre-cross-connected trails for optical mesh
network protection, OSA Journal of Optical
Networking, May 2006, pp.493-508
25High-Availability Network Architectures
(HAVANA)Comparative Study of Fully
Pre-Cross-Connected Protection Architectures for
Transparent Optical NetworksContact
grover_at_trlabs.ca
A. Grue, W. D. Grover, J. Doucette, B. Forst, D.
Onguetou, D. Baloukov TRLabs (Network Systems
Group) 7th Floor, 9107 116 Street Edmonton,
Alberta, Canada T6G 2V4
M. Clouqueur, D. Schupke Nokia Siemens
Networks (Network Control-Plane and
Transport) Otto-Hahn-Ring 6 81730 Munich, Germany
26Some Insights
- DSP
- - Why isn't it more efficient than it is ? (Turns
out almost identical to 11 APS) - - Amenability to exact design with ILP (design
ease) - PXTs
- High design and conceptual complexity
- Good flexibility for wavelength assignment,
optical path length constraints - p-Cycles
- Surprise that plain p-Cycles still have the best
spare capacity efficiency - Not inherently end-to-end path-protecting
- Optical Reach design control developed
- FIPP p-Cycles
- Offer a simple end-to-end protected path
tunnel operating paradigm - Exact ILP design possible, heuristics under
development
27Project HAVANA Ongoing Work
- Node Failure restorability analysis (and enhanced
design) - Detailed minimum-cost mapping of designs into
nodal equipment models - Costs associated with design for 100 node
failure restorability - Implications / feasibility of same wavelength
protection options in each architecture - Finding a good heuristic for FIPP p-Cycle design.
- Design for 100 R2 and/or to support multi-QoP
classes involving an ultra high availability
(R21) priority service.
28p-Trees / p-Cycles Computationally Distinct
p-Cycles Span p-Trees
PXTs/FIPP p-Cycles Path p-Trees
29The Z Case in FIPP p-Cycle Design
- Protection paths are pre-connected, but the
protection path to be used will depend on the
failure scenario - For the purpose of this study, the network was
deemed not intelligent enough to handle this
degree of failure dependency
30The Z Case in FIPP p-Cycle Design
- Protection paths are pre-connected, but the
protection path to be used will depend on the
failure scenario - For the purpose of this study, the network was
deemed not intelligent enough to handle this
degree of failure dependency
31Optical Path Lengths for p-Cycles
- In a path-protecting architecture, protection
paths are completely substituted for working
paths during failure, meaning that the lengths of
the restored state paths are not in question
- In a span-protecting architecture (p-Cycles, span
p-Trees), protection paths are only substituted
for the failed span, which may be used by many
working paths with different lengths