PhD Dissertation Defense - ITTC€¦ · PhD Dissertation Defense Service Profile-Aware Control Plane: A Multi-Instance Fixed Point Approximation within A Multi-Granularity VPN Loss

1

PhD Dissertation DefensePhD Dissertation DefenseService ProfileService Profile--Aware Control Plane:Aware Control Plane:

A MultiA Multi--Instance Fixed Point Approximation withinInstance Fixed Point Approximation withinA MultiA Multi--Granularity VPN Loss Networks PerspectiveGranularity VPN Loss Networks Perspective

Wesam Alanqar, PhDWesam Alanqar, PhDAdvisor: Prof. Victor FrostAdvisor: Prof. Victor Frost

April 25, 2005

2Outline

Problem statementIntroductionContributionsConfigured VPN service modelsControl plane models Analysis methodologyMathematical formulation highlightsScenarios and performance evaluation metricsMathematical models validationPerformance analysis results highlightsJustification for generalizing performance results for SPAConclusionsRecommendations Next steps and future workAppendixes:

– Appendix-A: Detailed mathematical formulation– Appendix-B: Detailed performance results (4-node topology)– Appendix-C: Detailed performance results (7-node topology- 2 alternate routing)– Appendix-D: Detailed performance results (7-node topology- 3 alternate routing)

3Problem Statement

The architectures and functional operation of existing control plane components do not consider the service profile layer parameters

This lack of harmony between the service profile layer, control plane layer, and network infrastructure layer exist in current IETF and ITU control plane models

This lack of harmony leads to inefficient utilization of network resources

Therefore, the problem is to develop a new Service Profile-Aware (SPA) control plane model that provides this harmony and then demonstrate its superiority over existing control plane models

SPA control plane components were architected to utilize both the service profile layer parameters and network infrastructure detailed resource representation parameters

4Introduction: Problem Motivation/Significance

This research proposes a new SPA control plane model – Detailed description of its architectural and functional operation– Analytically shows its superiority over existing IETF and ITU control plane models

The performance of the IETF/ ITU/SPA control plane models were analyzed in a common framework from the following perspectives:

– Transport network granularity realization– Operational level– Component-level interaction between the transport layer, control plane layer, and the service

profile layer

5Introduction: Research Approach

This research – Defined the architectures of the multiple configured VPN service proposed models– Defined the architectures and functional operation of the control plane components for the three

control plane models– Developed the mathematical models for the traffic management schemes of the three control

plane models• Used Fixed Point Approximation (FPA) analytical model to compute the performance

metrics for the traffic management schemes of the three control plane modelsThe superiority of the SPA control plane over IETF/ITU models was analytically demonstrated

6Contributions

Developed detailed architectures for the service configuration models from the following three perspectives:

– Service flow connectivity– Load partitioning flexibility– Service demand granularity

Developed architectures for the three control plane models (IETF, ITU, SPA) from the following three perspectives:

– Transport network granularity realization– Component-level– Operational-level

Developed mathematical formulation for each traffic management scheme of the three control plane models using Fixed Point Approximation (FPA)Demonstrated the superiority of the SPA over IETF and ITU control plane models using the following performance metrics:

– Service request blocking probability– Permissible “non-blocked” load– Transport resources utilization performance metrics

This work provides a significant shift in network design and traffic management for future wired and wireless networks

– More efficient utilization of network resources due to SPA enforcement of harmony between the service profile layer, control plane layer, and network infrastructure layer

7Configured VPN Service Models

Nine possible service models were considered based on the configured service profile parameters:

– Service flow connectivity (Point-Point, Semi-Meshed, Fully-Meshed)– Arrival load partitioning flexibility (Dedicated, Shared)– Service demand granularity (Granular, Coarse)

Configure VPN Service Models

Configure VPN Service Models

Point-to-Point

Point-to-Point

Semi-MeshedSemi-

MeshedFully-

MeshedFully-

MeshedLoad

partitioning disabled

Load partitioning

disabled

Load partitioning

enabled

Load partitioning

enabled

Granular Bandwidth Request


Coarse Bandwidth Request


Load partitioning

disabled

Load partitioning

disabled

Load partitioning

enabled

Load partitioning

enabled





Load partitioning

disabled

Load partitioning

disabled

Load partitioning

enabled

Load partitioning

enabled











Akb

Ckb

Gkb

Coarse Bandwidth

Actual Bandwidth

Akb

Gkb

Gkb

Akb

GranularBandwidth

Akb

Ckb

Gkb

Coarse Bandwidth

Actual Bandwidth

Akb

Gkb

Gkb

Akb

GranularBandwidth

8Configured VPN Service Models “Cont.”

Point-to-Point Dedicated Actual (PDA) Service Configuration

Fully-Meshed Dedicated Actual (FDA)Service Configuration

Point-to-Point Shared Actual (PSA) and Point-to-Point Shared Granular (PSG)

Service Configurations

Definitions and Notations

1. jC : The physical capacity or bandwidth of link j,

2. Dedicated ResourcevD

jC : The dedicated capacity on link j for configured service v.

3. Shared Resources SjC : The shared capacity on link j.

4. VPN Resources vjC : The VPN capacity on link j. S

jvDj

vj CCC +=

5. vrkλ : The arrival rate of class k calls between node pair r for configured service v.

6. vDrkλ : The dedicated arrival rate.

7. vSrkλ : The shared arrival rate.

8. Ckb : The coarse bandwidth requirement of class k calls

9. Gkb : The granular, sub-rate, bandwidth requirement of

Ckb , in units of bandwidth,

circuits

10. Akb : The actual bandwidth requirement of class k

9Configured VPN Service Models “Cont.”

Fully-Meshed Shared Actual (FSA) and Fully-Meshed Shared Granular (FSG)

Service Configurations

Definitions and Notations

1. jC : The physical capacity or bandwidth of link j,

2. Dedicated ResourcevD

jC : The dedicated capacity on link j for configured service v.

3. Shared Resources SjC : The shared capacity on link j.

4. VPN Resources vjC : The VPN capacity on link j. S

jvDj

vj CCC +=

5. vrkλ : The arrival rate of class k calls between node pair r for configured service v.

6. vDrkλ : The dedicated arrival rate.

7. vSrkλ : The shared arrival rate.

8. Ckb : The coarse bandwidth requirement of class k calls

9. Gkb : The granular, sub-rate, bandwidth requirement of

Ckb , in units of bandwidth,

circuits

10. Akb : The actual bandwidth requirement of class k

Analysis focused on Fully-Meshed Shared Granular (FSG) service model

– (point-to-point, semi-meshed) are subset of Fully-Meshed

– Dedicated arrival rate will not benefit from SPA Load Partitioning Function (LPF)

– Coarse service demands will not benefit from SPA Inverse Multiplexing Function (IMF)

10Control Plane Models- Traffic Management Schemes

Compared the three control plane models based on the following control plane traffic management capabilities:

– Routing update triggers (static vs. state-dependent)– Routing granularity level (coarse vs. granular)– Load Partitioning Function (LPF): Static Sharing (SS) vs. Network Engineering (NE)– Inverse Multiplexing Function (IMF): enabled vs. disabled

Routing granularity: Routing tables construction based transport granularity levelLoad Partitioning Function (LPF):

– Partition the service arrival load into two partitions; dedicated load and shared load– Has two options; Static Sharing (SS) and Network Engineering (NE)

• Static Sharing (SS): Statically partition configured service arrival load into two partitions Based on dedicated/shared resources to the VPN resources capacity ratio

• Network Engineering (NE): Dynamically partition arrival load between dedicated/shared resources

• Based on dedicated resources blocking probabilityInverse Multiplexing Function (IMF):

– Multiplexing/inverse multiplexing of incoming traffic– Multiplexing: Sending multiple signals or streams of information on a carrier at the same time– Inverse Multiplexing (IM): Dividing a data stream into multiple concurrent streams

• Transmitted at the same time across separate channels• Reconstructed at the other end back into the original data stream

11Control Plane Models- Traffic Management Schemes “Cont.”

IETF Control Plane Model

ITU Control Plane Model

SPA-Dedicated Control Plane ModelSPA-Shared Control Plane Model

12Control Plane Models- Traffic Management Schemes “Cont.”

EnabledEnabled (NE)State-DependentSPA- Shared W/ (NE,IM)

EnabledEnabled (SS)State-DependentSPA- Shared (W/O NE, W/ IM)

DisabledEnabled (NE)State-DependentSPA- Shared (W/ NE, W/O IM)

DisabledEnabled (SS)State-DependentSPA-Shared W/O (NE,IM)

DisabledDisabledState-DependentSPA-Dedicated

DisabledDisabledStatic- SplitITU-SR

DisabledDisabledStatic- DirectITU-DR

DisabledDisabledStatic- SplitIETF-SR

DisabledDisabledStatic- DirectIETF-DR

IMFLPFRouting ComponentControl Plane Model

Control Plane Model

Control Plane Model

IETFIETF ITUITU SPASPA

Dedicated VPNDedicated VPN Shared VPNShared VPN

Without Network Engineering (NE)

Without Network Engineering (NE)

With Network Engineering (NE)With Network

Engineering (NE)

Dedicated VPN

Dedicated VPN

Direct RoutingDirect Routing Split RoutingSplit Routing

Shared VPNShared VPN

Direct RoutingDirect RoutingSplit RoutingSplit Routing

Without Inverse Multiplexing (IM)Without Inverse

Multiplexing (IM)

With Inverse Multiplexing (IM)


Without Inverse Multiplexing (IM)Without Inverse

Multiplexing (IM)



13Control Plane Models- Transport Network Realization “Vertical View: Multi-Granularity”

Service Demand Granularity

Service Parameters

Service Configuration Profile


CkbCkb

Service Request with granular bandwidth Service request with

actual bandwidth

Akb Akb IMF Disabled

Multi-Granularity Transport

SNPSNP

Routing Area Routing Area

SNC

Coarse

representation of transport network granularity levels

Coarse



Service Parameters


ITUControl Plane Model

Akb Akb

Service Request with granular bandwidth

Service request with actual bandwidth

Akb Akb

IMF Disabled


SNPP.. ..SNPs SNPsSNP

SNPP..SNP SNPs SNP

Granular



SNCGranular representation of transport network granularity levels


Service Parameters


Service Profile-Aware Control Plane

Transport Network

Gkb Gkb

Service Request with Multiple granular bandwidth

Service request with actual bandwidth

Akb Akb

IMF Enabled


SNPP.. ..SNPs SNPsSNP

SNPP..SNP SNPs SNP

Granular representation of transport network granularity levels


SNCGranular representation of transport network granularity levels

Key Takeaway: From a transport resources granularity perspective:1. IETF control plane model has coarse resource representation2. ITU/SPA control plane models have granular resource representation

From a demand granularity perspective:1. IETF control plane model has coarse demand representation2. ITU control plane model has actual demand representation3. SPA control plane model have granular demand representation

Key Takeaway: From a transport resources granularity perspective:1. IETF control plane model has coarse resource representation2. ITU/SPA control plane models have granular resource representation

From a demand granularity perspective:1. IETF control plane model has coarse demand representation2. ITU control plane model has actual demand representation3. SPA control plane model have granular demand representation

14Control Plane Models- Transport Network Realization “Vertical View: Network Partitions”

One Control Plane Instance with one RDB for the Three Transport Network Partitions “VPNs”

LRMLRM


Network Partition “VPN-A”

Network Partition “VPN-C”

Network Partition “VPN-B”

RDBRDB

RCRC RCRC RCRCControl PlaneControl Plane

InstanceInstance

One Control Plane Instance with one RDB for the Three Transport Network Partitions “VPNs”

LRMLRM





RDBRDB

RCRC RCRC RCRCControl PlaneControl Plane

InstanceInstance

Three Control Plane Instance with Three RDB partitions for the Three Transport Network Partitions “VPNs”Without inter-control plane instances resource sharing via Load Partitioning Function (LPF)

LRM-ALRM-A

ITU/SPA-Dedicated Control Plane Models




RDB-ARDB-A

RCRC RCRC RCRC

LRM-CLRM-CRDB-CRDB-C

RCRC RCRC RCRC

LRM-BLRM-BRDB-BRDB-B

RCRC RCRC RCRC

Control PlaneControl PlaneInstanceInstance--AA

Control PlaneControl PlaneInstanceInstance--CC

Control PlaneControl PlaneInstanceInstance--BB

SPA-Shared Control Plane ModelThree Control Plane Instance with Three RDB partitions for the Three Transport Network Partitions “VPNs”

With inter-control plane instances resource sharing via Load Partitioning Function (LPF)

LRM-ALRM-A Network Partition “VPN-A”



RDB-ARDB-A

RCRC RCRC RCRC


RCRC RCRC RCRC


RCRC RCRC RCRC




Resource Sharing via LPF


SPA-Shared Control Plane ModelThree Control Plane Instance with Three RDB partitions for the Three Transport Network Partitions “VPNs”


LRM-ALRM-A Network Partition “VPN-A”



RDB-ARDB-A

RCRC RCRC RCRC


RCRC RCRC RCRC


RCRC RCRC RCRC






Key Takeaway: From a transport resources perspective:1. IETF control plane model implements Complete Sharing

(CS) concept2. ITU control plane model implements Complete Partitioning

(CP) concept3. SPA control plane model implements both CP and Virtual

Partitioning (VP) concepts

Key Takeaway: From a transport resources perspective:1. IETF control plane model implements Complete Sharing

(CS) concept2. ITU control plane model implements Complete Partitioning

(CP) concept3. SPA control plane model implements both CP and Virtual

Partitioning (VP) concepts

15Control Plane Models- Component-Level Interaction

Static Routing


Service Parameters

Transport Network

Path Computation

Coarse Transport Network Granularity Levels

Connection Setup


1

2

65

43


Routing Component

Service Flow Connectivity

Inverse Multiplexing Disabled 2

Inverse Multiplexing Function (IMF)

1

8

7

Static Routing


Configured VPN Service Identification Number

Service Parameters

Transport Network

Control Plane Instance (CPI) Selection

Path Computation

Granular Transport Network Granularity Levels

Connection Setup


2

4 87

65

9


Routing Component

2Inverse Multiplexing Disabled 3



111

State-DependentRouting



Service Parameters

Transport Network

Control Plane Instance (CPI) Selection

Path Computation


Connection Setup

SPA-DedicatedControl Plane Model

2

4 87

65

9


Routing Component

2Inverse Multiplexing Disabled 3



111

State-DependentRouting


Arrival RatePartitioning

Service Parameters

Transport Network

Service DemandGranularity

Inverse MultiplexingDisabled vs. Enabled

Network Engineering Vs. Static Sharing

Path Computation


Connection Setup

Current Topology Occupancy State “Per link”

SPA-SharedControl Plane Model

3 4

568

77

9

10

Service Configuration

Profile


Control Plane Instance (CPI) Selection1

2111

Routing Component

Inverse Multiplexing Function (IMF)Load Partitioning Function (LPF)

16Analysis Methodology- Fixed Point Approximation (FPA) Steps

Service Arrivals

)(np j

Estimating Link’s Reduced Load

Estimating Occupancy

Probabilities

jkamrkq

Estimating Admissibility Probabilities

Estimating Routing

Probabilities

In th

e ca

se o

f Sta

te-

Dep

ende

nt R

outin

g

],........,,[

],........,[

],.......,,[],......,,[

21

21

21

21

Kk

mrjk

mrj

mrj

mrjk

Kk bbbbKkkK

μμμμ

λλλλ

=

=

==

],....,,[],.....,,[

].,..........[log____#

].,,.........2,1[],........,2,1[

21

21

,2,1

jj

mm

rr

CCCCrrrr

MMMMytopoinpairsnodeofR

JJNN

==

====

2∑ ∏

∈

−=m mrj

jkmrkrk aqB 1 Compute Per-Route

Blocking Probability

Topology Parameters

1

jkλ

17Analysis Methodology- Assumptions

Link independence assumption – Blocking occurs independently from link to link, determined by their respective arrival rates– This assumption becomes more reasonable as traffic gets heavier

Poisson calls arrivals – The total offered load to an individual link is also a Poisson process with rate thinned by

blocking on other linksAll links are assumed to be undirectedStationary inputs

– Certain random quantities of interest have well-defined averages including:• Number of on-going calls on a link of each class• Average service request holding time• Reduced load on a link

– With these averages we can further assume that there is a stationary probability of choosing a particular route under the state-dependent routing scheme

Minimal route overlapping for analyzed topologies– FPA accuracy increases compared to DES

18Analysis Methodology- FPA for the Three Control Plane Models


)(np j



Probabilities

jkamrkq


Estimating Routing

Probabilities

One FPA Instance for the Three Transport Network Partitions “Dedicated Resources”

No State-Dependent Routing

Dedicated Resources-1



jkλEstimating Link’s

Reduced Load


Probabilities


Estimating Routing

Probabilities

Three FPA Instance for the Three Transport Network Partitions “Dedicated Resources”

No State-Dependent Routing





Per Network PartitionPer Network Partition

Per Network Partition


mDrkq

)(npDj

Djka

Djkλ



Probabilities


Estimating Routing

Probabilities

In th

e ca

se o

f Sta

te-

Dep

ende

nt R

outin

g

SPA-Dedicated Control Plane Model




Three FPA Instance for the Three Transport Network Partitions “Dedicated Resources”

With State-Dependent Routing




mDrkq

)(npDj

Djka

Djkλ Estimating Link’s

Reduced Load


Probabilities


Estimating Routing

Probabilities

In th

e ca

se o

f Sta

te-

Dep

ende

nt R

outin

g

SPA-Shared Control Plane Model


Shared Resources


Three FPA Instance for the Three Transport Network Partitions “2 Dedicated & 1 Shared Resources”

With State-Dependent Routing





mDrkq

)(npDj

Djka

Djkλ


19Analysis Methodology- FPA Modeling Environment

Occupancy Probability Computation

Occupancy Probability

Reduced Load Computation

Visual BasicTM for FPA

Routing Probability Computation

MathematicaTM

Code

20Mathematical Formulation Highlights “Details available in Appendix-A”

Base FPA method was modified based on the unique attributes of each traffic management scheme of the three control plane modelsMathematical formulation steps included:

– CAC for multi-rate demands• Modified based on each control plane representation of demand granularity

– Estimating link’s reduced load• Modified based on each control plane input load handling capability

– Estimating link’s admissibility probability• Modified based on each control plane representation of transport resources granularity

– Estimating routing probability• Modified based on each control plane routing static vs. state-dependent mechanism

– Estimating network blocking probability• Modified based on the each control plane representation of network partitions

– Estimating network permissible load• Modified based on the each control plane representation of network partitions

– Estimating network utilization• Modified based on the each control plane representation of network partitions

21Scenarios & Performance Evaluation- Control Plane Parameters & Performance Metrics

22Scenarios & Performance Evaluation- FPA modeling parameters

Service Arrivals

2

∑ ∏∈

−=m mrj

jkmrkrk aqB 1 Compute Per-Route

Blocking Probability

Topology Parameters

1

)(np j



Probabilities

jkamrkq


Estimating Routing

Probabilities

In th

e ca

se o

f Sta

te-

Dep

ende

nt R

outin

g

jkλ

)(np j



Probabilities

jkamrkq


Estimating Routing

Probabilities

In th

e ca

se o

f Sta

te-

Dep

ende

nt R

outin

g

jkλ

]1[

]3010[

]12[]2[

1 ==

>−==

−===

μμ

λλ

k

rjk

rjk

k

Erlang

STSbKK

mm

]24,24,24,24,24,24[}]4,3{},4,2{},3,2{},4,1{},3,1{},2,1{[

]2,2,2,2,2,2[6

]4,3,2,1[]4,3,2,1[

654321

654321

654321

==============

========

==========

CCCCCCCrrrrrrr

MMMMMMMR

jjjjJnnnnN

j

m

r

n=1 n=2

n=3n=1

j=1

j=2

j=3

j=4

23Scenarios & Performance Evaluation- Network Topologies Analyzed

Physical resources in:IETF Control Plane Model

Network resources partitions in:ITU ,SPA-Dedicated

DedicatedResource-1

DedicatedResource-2

n=1 n=2

n=3n=1

n=1 n=2

n=3n=1

n=1 n=2

n=3n=1

j=1

j=2

j=3

j=4

j=1

j=2

j=3

j=4

j=1

j=2

j=3

j=4

124 −= STSC j

112 −= STSC vDj

112 −= STSC vDj


Network resources partitions in:ITU ,SPA-Dedicated

DedicatedResource-1

DedicatedResource-2

n=1 n=2

n=3n=1

n=1 n=2

n=3n=1

n=1 n=2

n=3n=1

j=1

j=2

j=3

j=4

j=1

j=2

j=3

j=4

j=1

j=2

j=3

j=4

124 −= STSC j

112 −= STSC vDj

112 −= STSC vDj


Network resources partitions in:SPA-Shared

Resource Resource SharingSharing


DedicatedResource-1

SharedResources

DedicatedResource-2

n=1 n=2

n=3n=1

j=1

j=2

j=3

j=4

n=1 n=2

n=3n=1

j=1

j=2

j=3

j=4

n=1 n=2

n=3n=1

j=1

j=2

j=3

j=4

n=1 n=2

n=3n=1

j=1

j=2

j=3

j=4

124 −= STSC j

111 −= STSC vDj

111 −= STSC vDj

12 −= STSC Sj


Network resources partitions in:SPA-Shared



DedicatedResource-1

SharedResources

DedicatedResource-2

n=1 n=2

n=3n=1

j=1

j=2

j=3

j=4

n=1 n=2

n=3n=1

j=1

j=2

j=3

j=4

n=1 n=2

n=3n=1

j=1

j=2

j=3

j=4

n=1 n=2

n=3n=1

j=1

j=2

j=3

j=4

124 −= STSC j

111 −= STSC vDj

111 −= STSC vDj

12 −= STSC Sj

DedicatedResource-1

DedicatedResources-2

Physical resources in:IETF Control Plane

Model

Network resources partitions in: ITU ,SPA-Dedicated

112 −= STSC vDj

112 −= STSC vDj

124 −= STSC j

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

DedicatedResource-1

DedicatedResources-2


Model

Network resources partitions in: ITU ,SPA-Dedicated

112 −= STSC vDj

112 −= STSC vDj

124 −= STSC j

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

Network resources partitions in: SPA-Shared

DedicatedResource-1

SharedResources

DedicatedResource-2



Model


124 −= STSC j 12 −= STSC Sj

111 −= STSC vDj

111 −= STSC vDj

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

Network resources partitions in: SPA-Shared

DedicatedResource-1

SharedResources

DedicatedResource-2



Model



111 −= STSC vDj

111 −= STSC vDj

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

DedicatedResource-1

SharedResources

DedicatedResource-2



Model



111 −= STSC vDj

111 −= STSC vDj

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

j=1 j=2 j=3

j=4

j=5j=6

j=7 j=8 j=9

n=1n=2 n=3 n=4

n=5n=6n=7

24Mathematical Models Validation- Credibility of Modeling Results

Accuracy of mathematical models assumptions– Higher input loads were used to improve the accuracy of results

• The base method validated that for blocking probabilities under higher input loads, FPA algorithm average percentage error compared to DES is below 5%

– Minimal route overlapping was considered for the 4-node and 7-node topologies analyzed• The base method validated that under minimal routing overlapping for fully connected and

random topologies, FPA algorithm accuracy compared to DES increases Accuracy of occupancy probability computation

– When the output of the of the occupancy probabilities equations was used in the FPA algorithm, validating that the summation of occupancy probabilities of link j for all the states is equal to 1 was carried after each FPA convergence, the percentage of error was 0%

Accuracy of routing probability computation– After each FPA convergence, the routing probability constraint, summation of the routing

probability for all the routes between a source-destination pair has to equal 1, was validated– Percentage error was in the range below 3% for the 7-node topology and 0% for the 4-node

topologyAccuracy of LPF and IMF traffic management operations

– LPF sanity check: made sure that the summation of the load applied to the dedicated network resources partitions and the shared network resources partition is equal to the total input load

– IMF sanity check: made sure that the input load before an inverse multiplexing operation is equal to the input load after the inverse multiplexing operation

25Mathematical Models Validation- Performance Results Consistent Trends “7-node topology with both 2 & 3 alternate routing”

Blocking Probability:– SPA-Dedicated traffic management scheme does not provide any reduction in blocking

probability compared to the IETF-DR traffic management scheme

• But provides higher reduction in blocking probability compared to IETF-SR and ITU-SR traffic management schemes

– The SPA two traffic management schemes with enabled inverse multiplexing lead to the highest reduction in blocking probability, compared to the rest of the traffic management schemes

Permissible Load:– The SPA two traffic management schemes with enabled inverse multiplexing lead to the highest

increase in permissible load compared to the rest of the traffic management schemes

– The SPA-Shared control plane model with disabled inverse multiplexing leads to a reduction in permissible load compared to the IETF-DR traffic management scheme

Utilization:– SPA-Shared with enabled inverse multiplexing leads to a lower reduction in utilization

compared to the IETF-DR traffic management scheme

– SPA-Shared with disabled inverse multiplexing leads to higher reduction in utilization compared to IETF-DR traffic management scheme

26Performance Analysis Results Highlights- Blocking Probability “Summary View”

Network-Wide Reduction in Blocking Probability (Physical Resources Level)- IETF-DR as reference control plnae model

7-Node Topology (3- Alternate Routing)

0%

-43%

9%

-35%

0%

-9%-5%

26%

43%

-50%

-40%

-30%

-20%

-10%

0%

10%

20%

30%

40%

50%

IETF-DR IETF-SR ITU-DR ITU-SR SPA-Dedicated

SPA-w/o(NE,IM)

SPA-w/NE,w/oIM

SPA-w/oNE,w/IM

SPA-w/(NE,IM)

Traffic Management Scheme

% R

educ

tion

in B

lock

ing

Prob

abili

ty

All SPA traffic management schemes provide a higher reduction in blocking probability compared to the IETF-SR and ITU-SR control plane models

– Reduction is 0-131% and 39-122% respectively; depending on the SPA traffic management scheme, and SPA number of alternate routes

When IMF is disabled, IETF-DR traffic management scheme produces less blocking probability than SPA control plane modelWhen IMF is enabled, SPA control plane model leads to the highest reduction in blocking probability compared to IETF-DR

– Reduction of 22-48% depending on the number of alternate routes

27Performance Analysis Results Highlights- Permissible Load “Summary View”

Network-Wide Increase in Permissible Load (Physical Resources Level)- IETF-DR as reference control plnae model


0% 0% 0% 1% 2%

-1%-11%

120%111%

-20%

0%

20%

40%

60%

80%

100%

120%

140%


SPA-w/o(NE,IM)

SPA-w/NE,w/oIM

SPA-w/oNE,w/IM

SPA-w/(NE,IM)


% R

educ

tion

in B

lock

ing

Prob

abili

ty

All SPA traffic management schemes, except when IMF is disabled, provide a higher increase in permissible load compared to the IETF-SR and ITU-(DR,SR) control plane models

– Increase is 120-134% and 110-120% respectively; depending on the SPA traffic management scheme, SPA number of alternate routes, and the IETF/ITU static routing configuration

Highest increase in permissible load occurs for SPA-“w/oNE,w/IM” traffic management scheme– Increase is 120-134% compared to IETF-DR control plane model; depending on the number of alternate routes

While enabling IM, performing load partitioning statically or dynamically does not provide a significant impact on the percentage gain in permissible loadWhile disabling IM and regardless of static or dynamic load partitioning for SPA-Shared control plane model, the SPA control plane model provides less permissible load than IETF-DR control plane model

28Performance Analysis Results Highlights- Utilization “Summary View”

Network-Wide Reduction in Utilization (Physical Resources Level)- IETF-DR as reference control plnae model


0%

-14%

13%

-3%

1%

19%21%

-6%

-8%

-20%

-15%

-10%

-5%

0%

5%

10%

15%

20%

25%


SPA-w/o(NE,IM)

SPA-w/NE,w/oIM

SPA-w/oNE,w/IM

SPA-w/(NE,IM)


% R

educ

tion

in B

lock

ing

Prob

abili

ty

All SPA traffic management schemes, except when IMF is enabled, provide a higher reduction in utilization compared to the ITU-(DR,SR) control plane models

– Reduction is 8-28% depending on the SPA traffic management scheme, number of SPA alternate routes, and the ITU static routing configuration

SPA control plane model provide a reduction in utilization only when IMF is disabled– Reduction is 19-23% depending on the SPA traffic management scheme and number of alternate routes

The lowest reduction in utilization occur for “w/(NE,IM)” and “w/oNE,w/IM” SPA traffic management schemes when IMF is configured to enabled Inverse Multiplexing (with IM) and regardless of LPF configuration as static or dynamic partitioning

– Increase of 2-8% in utilization over the IETF-DR control plane model depending on the number of alternate routes

29Justification for Generalizing Performance Analysis Results for SPA

State-dependent routing distributes the input load across all the identified routes between a source-destination pair based on the traffic occupancy rather than static routing as in IETF/ITU control plane models

– This leads to: Blocking probability reduction, permissible load slight increase, and utilization reduction

LPF utilizes both the dedicated resources partition and the shared resources partition for the configured VPN service

– Thus, the configured VPN service will have more resources than IETF and ITU control plane models

– This leads to: Blocking probability reduction and utilization reduction

IMF splits incoming service request flows between a source-destination pair with an actual bandwidth requirement into multiple flows each with granular bandwidth requirement

– Each granular flow is routed independently across the available routes– This leads to: Blocking probability reduction, permissible load increase, and utilization increase

30Conclusions

All SPA traffic management schemes provide a higher reduction in blocking probability compared to the IETF-SR and ITU-SR control plane models

– Reduction is 0-131% and 39-122% respectively When IMF is enabled, SPA control plane model leads to the highest reduction in blocking probability compared to IETF-DRAll SPA traffic management schemes, except when IMF is disabled, provide a higher increase in permissible load compared to the IETF and ITU control plane models

– Increase is 120-134% and 110-120% respectivelySPA-Shared with enabled load sharing and disabled inverse multiplexing provide a higher reduction in utilization compared to all IETF and ITU traffic management schemes

– Reduction is 8-35% The performance analysis results carried on the 7-node topologies for both two and three alternate routes:

– Validated the hypotheses of this work– Indicated a common trend of the superiority of the SPA control plane model over the IETF and

ITU control plane modelsThus, the performance analysis concluded with SPA superiority over existing IETF/ITU control plane models

– SPA provides a significant shift in network design and traffic management for future wired and wireless networks

– More efficient utilization of network resources due to SPA enforcement of harmony between the service profile layer, control plane layer, and network infrastructure layer

31Recommendations

To achieve maximum blocking probability reduction over IETF and ITU control plane models, SPA components need to be configured as follows:

– State-dependent routing: Enabled– Inverse Multiplexing Function (IMF): Enabled– Load Partitioning Function (LPF): configured as Network Engineering (NE)

To achieve maximum permissible load increase over IETF and ITU control plane models, SPA components need to be configured as follows:

– State-dependent routing: Enabled– Inverse Multiplexing Function (IMF): Enabled– Load Partitioning Function (LPF): configured as Static Sharing (SS) or Network Engineering

(NE)To achieve maximum reduction in utilization over IETF and ITU control plane models, SPA components need to be configured as follows:

– State-dependent routing: Enabled– Inverse Multiplexing Function (IMF): Disabled– Load Partitioning Function (LPF): configured as Static Sharing (SS)

32Next Steps/Future Work- Multi-Domain Analysis

Core Backbone

Metro

Access

Resource Ownership

Traffic Locality

Network Granularity

Core Backbone

Metro

Access

Resource Ownership

Traffic Locality

Network Granularity

Dissertation Scope

FutureWork

Develop methods to predict the performance of the three control plane models for larger topologiesHierarchal routing architecture is needed

– To overcome the current limitations of the routing probability approximation – Current FPA mechanism lacks accuracy under large network topologies with flat routing

architecture

33Next Steps/Future Work- Multi-Domain Analysis “Hierarchal Routing”

One Control Plane Instance with one Hierarchal Routing Instance for the Three Transport Network PartitionsIETF Control Plane Model

One Control Plane InstanceOne Control Plane Instance




Three Control Plane Instance with Three Hierarchal Routing Instances for the Three Transport Network PartitionsNo inter-control plane instances resource sharing via Load Partitioning Function (LPF)

ITU/SPA-Dedicated Control Plane Models




SPA-Shared Control Plane ModelThree Control Plane Instance with Three Hierarchal Routing Instances for the Three Transport Network Partitions


CPICPI--AA

CPICPI--CC

CPICPI--BB

Resource Sharing via LPFResource Sharing via LPF

Resource Sharing via LPFResource Sharing via LPF




34

Discussion & Questions!Discussion & Questions!

PhD Candidate: Wesam AlanqarPhD Candidate: Wesam AlanqarAdvisor: Prof. Victor FrostAdvisor: Prof. Victor Frost

April 25, 2005

35Appendix-A: Mathematical Formulation- Step-1: CAC for Multi-Rate Demands

IETF control plane model– Routing component advertises the traffic occupancy of the coarse granularity levels– Service request with actual bandwidth requirements ( ) will consume ( ) resources from ( ) – A call with bandwidth requirement ( ) will be accepted if the following condition apply:

ITU & SPA-Dedicated control plane models– Routing component advertises the traffic occupancy of the fine granularity levels– Service request with actual bandwidth requirements ( ) will consume ( ) resources from ( ) – A call with bandwidth requirement ( ) will be accepted if the following condition apply:

SPA-Shared control plane model– Differs from both the ITU and SPA-Dedicated control plane models

• Can enable IMF and further divide the service request demand ( )into sub-rates or granular demands ( )

– A call with bandwidth requirement ( ) will be accepted if the following condition apply:

Akb C

kb jCAkb

∑∈

−≤Kk

kj

Ckj

Ak nbCb

Akb A

kb jCAkb

∑∈

−≤Kk

Djk

Ak

Dj

Ak nbCb

Akb

Gkb

Akb

∑∈

−≤Kk

vDjk

Gk

vDj

Gk nbCb ∑

∈

−≤Kk

vSjk

Gk

vSj

Gk nbCbDedicated

Resources:Shared

Resources:

36Appendix-A: Mathematical Formulation- Step-2: Estimating Link’s Reduced Load

IETF control plane model

ITU & SPA-Dedicated control plane models

SPA-Shared control plane model (Without NE)

∏≠∈

∈=jiri

ikmmrk

vrkjk

m

mr arjIq,

][λλ ∑ ∑∈ ∈

=Rr Mr

jkjkrm

mrλλ

∏≠∈

∈=jiri

Dikm

mDrk

Drk

Djk

m

rm arjIq,

][λλ ∑ ∑∈ ∈

=Rr Mr

Djk

Djk

rm

mrλλ

vSj

vDj

vDjv

rkvDrk CC

C+

= .λλ

vSj

vDj

vSjv

rkvSrk CC

C+

= .λλ

∏≠∈

∈=jimri

Dikm

mDrk

vDrk

mrDjk arjIq

,][λλ ∑ ∑

∈ ∈

=Rr Mr

Djk

Djk

rm

mrλλ

)(~ ∑∀

=v

vSrk

Srk λλ ∏

≠∈

∈=jiri

Sikm

Smrk

Srk

Sjk

m

mr arjIq,

][~λλ ∑ ∑∈ ∈

=Rr Mr

Sjk

Sjk

rm

mrλλ

SPA-Shared control plane model (With NE)

∏≠∈

∈=jimri

Dikm

mDrk

vrk

mrDjk

NE arjIq,

][λλ ∑ ∑∈ ∈

=Rr Mr

Djk

NEDjk

NE

rm

mrλλ vDrk

vrk

vSrk

NE B.λλ = ∑ ∏∈

−=m mrj

vDjk

mDrk

vDrk aqB 1

)(~ ∑∀

=v

vSrk

Srk

NE λλ ∏≠∈

∈=jiri

Sikm

Smrk

Srk

NESjk

NE

m

mr arjIq,

][~λλ ∑ ∑∈ ∈

=Rr Mr

Sjk

NESjk

NE

rm

mrλλ

)1.( Drk

vrk

vDrk

NE B−= λλ ∏≠∈

∈=jiri

Dikm

Dmrk

vrk

NEDjk

NE

m

mr arjIq,

][λλ

Roun

d-1

Roun

d-2

Round-1“Dedicated Resources Only”

Link j reduced loadbased on class k

Dedicated resources partition D reduced load on Link j based on class k

Dedicated resources partition D reduced load on Link j based on class k

37Appendix-A: Mathematical Formulation- Step-3: Estimating Link’s Admissibility Probability



SPA-Shared control plane model “without (NE,IM)”

∑ −=K

Ckj

k

jkCkj bnpbnnp )()(μλ

∑ −

==

Ckj bC

n jjk npa0

)(

∑ −=vS

Ak

Dj

k

DjkA

kDj bnpbnnp )()(

μλ ∑ −

==

Ak

Dj bC

nDj

Djk npa

0)(

j

C

Dj

Djk

jk C

Ca

aDj

∑∀

=

*

∑ −=K

Ak

vDj

k

vDjkA

kvDj bnpbnnp )()(

μλ

∑−

=

=

AkbvD

jC

n

vDj

vDjk npa

0)(

∑ −=K

Ak

Sj

k

SjkA

kSj bnpbnnp )()(

μλ

∑−

=

=Ak

Sj bC

n

Sj

Sjk npa

0)(

Sj

vDj

Sj

Sjk

vDj

vDjkv

jk CCCaCa

a+

+=

..

j

Sj

Sjk

D

vDj

vDjk

jk C

CaCaa

.).( +=∑∀

SPA-Shared control plane model “with/NE, without/IM”

∑ −=K

Ak

vDj

k

Djk

NEAk

vDj bnpbnnp )()(

μλ

∑−

=

=

AkbvD

jC

n

vDj

vDjk npa

0)(

∑ −=K

Ak

Sj

k

Sjk

NEAk

Sj bnpbnnp )()(

μλ

∑−

=

=Ak

Sj bC

n

Sj

Sjk npa

0)(

Sj

vDj

Sj

Sjk

Dj

vDjkv

jk CCCaCa

a+

+=

..j

Sj

Sjk

D

Dj

Djk

jk C

CaCaa

.).( +=∑∀

Dedicated resources D of VPN v admissibility probabilityon link j for class k

Shared resources S admissibility probability on link j for class k

Link j admissibility probability based on class k



38Appendix-A: Mathematical Formulation- Step-3: Estimating Link’s Admissibility Probability “Cont.”

SPA-Shared control plane model “without/NE, with/IM”

SPA-Shared control plane model “with/(NE,IM)”

∑ −=K

Gk

vDj

k

DjkG

kvDj bnp

ibnnp )()(

μλ

∑−

=

=

GkbvD

jC

n

vDj

vDjk npa

0)(

∑ −=K

Gk

Sj

k

SjkG

kSj bnp

ibnnp )()(

μλ ∑

−

=

=Gk

Sj bC

n

Sj

Sjk npa

0)(

Sj

vDj

Sj

Sjk

Dj

vDjkv

jk CCCaCa

a+

+=

..

j

Sj

Sjk

D

Dj

Djk

jk C

CaCaa

.).( +=∑∀

∑ −=K

Gk

vDj

k

Djk

NEGk

vDj bnp

ibnnp )()(

μλ

∑ −=K

Gk

Sj

k

Sjk

NEGk

Sj bnp

ibnnp )()(

μλ

∑−

=

=Gk

Sj bC

n

Sj

Sjk npa

0)(

Sj

vDj

Sj

Sjk

Dj

vDjkv

jk CCCaCa

a+

+=

..

j

Sj

Sjk

D

Dj

Djk

jk C

CaCaa

.).( +=∑∀

Dedicated resources partition D of VPN v admissibility probability on link j for class k

Shared resources partition S admissibility probability on link j for class k


∑−

=

=Gk

vDj bC

n

vDj

vDjk npa

0)(

39Appendix-A: Mathematical Formulation- Step-4: Estimating Routing Probability

∏∑∈

−

=

=)( 0

)()](Pr[m

Dj

rj

nC

k

Djm

Dn kPrA

∏ ∑∈

+−

=+ =

)(

1

01 )()](Pr[

m

Dj

rj

nC

k

Djm

Dn kPrA

∏ ∑−∈

−

=

=−)( 0

)()](Pr[mk

Dj

rrj

nC

k

Djmk

Dn kPrrA ∏ ∑

−∈

−

=

−=−)( 0

)(1)](Pr[mk

Dj

rrj

nC

k

Djmk

Dn kPrrA

∏ ∑−∈

+−

=+ =−

)(

1

01 )()](Pr[

mk

Dj

rrj

nC

k

Djmk

Dn kPrrA ∏ ∑

−∈

+−

=+ −=−

)(

1

01 )(1)](Pr[

mk

Dj

rrj

nC

k

Djmk

Dn kPrrA

)](Pr[)](Pr[)](~Pr[ 1 mDnm

Dnm

Dn rArArA +−=

∑ ∏∏=

+

=

+=

−=

=

−−=)(

01

1

1

1

min

)](~Pr[)].(Pr[)].(Pr[m rrC

nm

Dnmk

Dn

Mk

mkmk

Dn

mk

k

mDrk rArrArrAq

SPA-Shared control plane model– Estimating routing probability is carried independently for both dedicated and shared resources

partitions • Similar set of equations like SPA-dedicated but with different notations

SPA-Dedicated control plane model– Estimating routing probability is carried independently for each dedicated resources partition

Dedicated resources partition D of VPN v routing probability on pair r for class k

40Appendix-A: Mathematical Formulation- Step-5: Estimating Network Blocking Probability


SPA-Shared control plane models

∑ ∏∈

−=m rj

jkmrkrk

m

aqB 1 ][ rkRrk BAVRB∈

=


∑ ∏∈

−=m rj

Djk

Dmrk

Drk

m

aqB 1 ][ DrkRr

Dk BAVRB

∈=

j

C

Dj

Drk

rk C

CBB

Dj

∑∀=

*

][ rkRrk BAVRB∈

=

∑ ∏∈

−=m mrj

Djk

mDrk

Drk aqB 1 ][ D

rkRr

Dk BAVRB

∈=

∑ ∏∈

−=m mrj

Sjk

mSrk

Srk aqB 1 ][ S

rkRr

Sk BAVRB

∈=

Sj

Dj

Sj

Srk

Dj

Drkv

rk CCCBCB

B+

+=

** ][ vrkRr

vk BAVRB

∈=

j

Sj

Srk

DjC

Dj

Drk

rk C

CBCB

B

*)*( +

=

∑∀

][ rkRrk BAVRB∈

=

Network-wide blocking probability for class k

Dedicated resources partition D network-wide blocking probability for class k

Shared resources partition S network-wide blocking probability for class k

VPN resources partition v network-wide blocking probability for class k

Network-wide blocking probability for class k

41Appendix-A: Mathematical Formulation- Step-6: Estimating Network Permissible Load




)(ˆ1

jk

M

m rj

mrkrk

r

m

MINq λλ ∑= ∈

= ]ˆ[ˆrkRrk Avr λλ

∈=

)(ˆ1

Djk

M

m rj

mDrk

Drk

r

m

MINq λλ ∑=

∈=

j

v

Dj

Drk

rk C

C )*ˆ(ˆ

∑∀=λ

λ

)(ˆ1

Djk

M

m rj

mDrk

Drk

r

m

MINq λλ ∑=

∈= ]ˆ[ˆ D

rkRr

Dk Avr λλ

∈=

j

v

Dj

Drk

rk C

C )*ˆ(ˆ

∑∀=λ

λ

)(ˆ1

Sjk

rM

m mrj

mSrk

Srk MINq λλ ∑

= ∈= ]ˆ[ˆ S

rkRr

Sk Avr λλ

∈=

Sj

Dj

Sj

Srk

Dj

Drkv

rk CCCC

+

+=

*ˆ*ˆˆ λλλ

j

Sj

Srk

D

Dj

Drk

rk C

CC *ˆ)*ˆ(ˆ

λλλ

+=∑∀

Network-wide permissible “non-blocked” load for class k

VPN resources partition v network-wide permissible “non-blocked” load for class k

Shared resources partition s network-wide permissible “non-blocked” load for class k

Dedicated resources partition D network-wide permissible “non-blocked”load for class k

42Appendix-A: Mathematical Formulation- Step-7: Estimating Network Utilization




∑=

=jC

njj npU

1)( ][ jJj

UAvrU∈

=

∑=

=DjC

n

Dj

Dj npU

1

)(j

v

Dj

Dj

j C

CUU

)*(∑∀= ][ D

jJj

D UAvrU∈

=

∑=

=DjC

n

Dj

Dj npU

1

)(

∑=

=

SjC

n

Sj

Sj npU

1)(

Sj

Dj

Sj

Sj

Dj

Djv

j CCCUCU

U+

+=

**

j

Sj

Sj

v

Dj

Dj

j C

CUCUU

*)*( +=∑∀

][ jJjUAvrU

∈=

Network-wide utilization

Network-wide utilization of dedicated network resources partition D

Network-wide utilization of shared network resources partition S for link j

Network-wide utilization of VPN network resources partition v for link j

Link j utilization based on the utilization of dedicated and shared network resource partitions

43Appendix-B: 4 Node Topology Detailed Results “Blocking probability”

4-node Topology (Fully-meshed Service Configuration)Average Network-Wide Blocking Probability (Physcial Resources)

2-Alternate Routing, Class-B Arrivals, IETF(DR,SR), ITU(DR,SR), SPA-Dedicated

0 5 10 15 20 25 30 35 40

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

Blo

ckin

g Pr

obab

ility


Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. SPA- Dedicated2. ITU-SR3. ITU-DR4. IETF-SR5. IETF-DR

Blocking Key Takeaways: SPA-Dedicated leads to lower blocking probability than IETF & ITU control plane models.


4-node Topology (Fully-meshed Service Configuration)Average Network-Wide Blocking Probability (Physical Resources)

2-Alternate Routing, Class-B Arrivals, IETF(DR,SR), ITU(DR,SR), SPA-w/o(NE,IM)

0 5 10 15 20 25 30 35 40

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

Blo

ckin

g Pr

obab

ility

SPA-w/o(NE,IM )-4S ITU-DR ITU-SR IETF-DRIETF-SR SPA-w/o(NE,IM )-1S SPA-w/o(NE,IM )-2S SPA-w/o(NE,IM )-3S

Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. SPA- w/o(NE,/IM)-4S2. SPA- w/o(NE,/IM)-3S3. SPA- w/o(NE,/IM)-1S4. SPA- w/o(NE,/IM)-2S5. ITU-SR6. ITU-DR7. IETF-SR8. IETF-DR

Blocking Key Takeaways: 1. At higher input loads, disabling NE & IM under any sharing ratio leads to lower blocking probability than both IETF and ITU models.2. Increasing sharing ratio leads to lower blocking probability on the SPA-w/o(NE,IM)



2-Alternate Routing, Class-B Arrivals, IETF(DR,SR), ITU(DR,SR), SPA-(w/NE,w/oIM)

0 5 10 15 20 25 30 35 40

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

Blo

ckin

g Pr

obab

ility

SPA-(w/ NE,w/oIM)-4S ITU-DR ITU-SR IETF-DRIETF-SR SPA-(w/ NE,w/oIM)-1 S SPA-(w/ NE,w/oIM)-2S SPA-(w/ NE,w/oIM)-3S

Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. SPA- (w/NE,w/o/IM)-1S2. SPA- (w/NE,w/o/IM)-3S3. SPA- (w/NE,w/o/IM)-2S4. SPA- (w/NE,w/o/IM)-4S5. ITU-SR6. ITU-DR7. IETF-SR8. IETF-DR

Blocking Key Takeaways: 1. Under high input loads, enabling NE only leads to lower blocking probability than IETFand ITU control plane models2. Increasing sharing ratio has no direct effect on blocking probability on the SPA-(w/NE,w/oIM)



2-Alternate Routing, Class-B Arrivals, IETF(DR,SR), ITU(DR,SR), SPA-(w/oNE,w/IM)

0 5 10 15 20 25 30 35 40

Input Load (Erlang)

-0.05

0.00

0.05

0.10

0.15

0.20

0.25

0.30

Blo

ckin

g Pr

obab

ility

ITU-DR ITU-SR IETF-DR IETF-SRSPA-(w/o NE,w/IM)-1S SPA-(w/o NE,w/IM)-2S SPA-(w/o NE,w/IM)-3S SPA-(w/o NE,w/IM)-4S

Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. SPA- (w/oNE,w//IM)-4S2. SPA- (w/oNE,w//IM)-3S3. SPA- (w/oNE,w//IM)-2S4. SPA- (w/oNE,w//IM)-1S5. ITU-SR6. ITU-DR7. IETF-SR8. IETF-DR

Blocking Key Takeaways: 1. Enabling IM under any sharing ratio leads to lower blocking probability than IETF & ITU control plane models2. Increasing sharing ratio leads to lower bocking probability on the SPA-(w/oNE,w/IM)



2-Alternate Routing, Class-B Arrivals, IETF(DR,SR), ITU(DR,SR), SPA-w/(NE,IM)

0 5 10 15 20 25 30 35 40

Input Load (Erlang)

-0.05

0.00

0.05

0.10

0.15

0.20

0.25

0.30

Blo

ckin

g Pr

obab

ility

SPA-(w/ NE,IM)-4S ITU-DR ITU-SR IETF-DRIETF-SR SPA-(w/ NE,IM)-1S SPA-(w/ NE,IM)-2S SPA-(w/ NE,IM)-3S

Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. SPA- w/(NE,/IM)-1S2. SPA- w/(NE,/IM)-2S3. SPA- w/(NE,/IM)-3S4. SPA- w/(NE,/IM)-4S5. ITU-SR6. ITU-DR7. IETF-SR8. IETF-DR

Blocking Key Takeaways: 1. Enabling both NE & IM under any sharing ratio leads to lower blocking probability than IETF-DR, ITU-DR, IETF-SR, and ITU-SR2. Enabling NE in addition to IM leads to lower blocking probability.3. Increasing sharing ratio leads to higher bocking probability on the SPA-w/(NE,IM)

48Appendix-B: 4 Node Topology Detailed Results “Permissible Load”

4-node Topology (Fully-meshed Service Configuration)Average Network-Wide Permissible Load (Physical Resources)


0 5 10 15 20 25 30 35 40

Input Load (Erlang)

-1.00

1.00

3.00

5.00

7.00

9.00

11.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. SPA-Dedicated2. IETF-DR3. ITU-DR4. IETF-SR5. ITU-SRUnder any given input load:1. No significant permissible load advantage of the SPA-Dedicated over both IETF-DR and ITU-DR2. ITU-SR& IETF-SR provides a higher permissible load than (IETF-DR,ITU-DR, and SPA-Dedicated

Permissible load Key Takeaways: 1. Split routing provides higher permissible load than direct routing for both IETF and ITU control plane models.




0 5 10 15 20 25 30 35 40

Input Load (Erlang)

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. SPA-w/o(NE,IM)-2S2. SPA-w/o(NE,IM)-4S3. SPA-w/o(NE,IM)-3S4. SPA-w/o(NE,IM)-1S5. IETF-DR6. ITU-DR7. IETF-SR8. ITU-SR

Permissible load Key Takeaways: 1. SPA-w/o(NE,IM) provides lower permissible load compred to IETF and ITU models under both direct and split routing.




0 5 10 15 20 25 30 35 40

Input Load (Erlang)

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

Per-

Pair

Perm

issi

ble

Load

SPA-(w/ NE,w/oIM)-4S ITU-DR ITU-SR IETF-SRIETF-SR SPA-(w/ NE,w/oIM)-1 S SPA-(w/ NE,w/oIM)-2S SPA-(w/ NE,w/oIM)-3S

Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. SPA-(w/NE,w/oIM)-4S2. SPA-(w/NE,w/oIM)-2S3. SPA-(w/NE,w/oIM)-3S4. SPA-(w/NE,w/oIM)-1S5. IETF-DR6. ITU-DR7. IETF-SR8. ITU-SRPermissible load Key Takeaways: 1. SPA-(w/NE,w/oIM) provides a lower permissible load compred to IETF and ITU models under both direct and split routing.2. For SPA-(w/NE,w/oIM) model, increasing sharing ratio leads to lower permissible load.




0 5 10 15 20 25 30 35 40

Input Load (Erlang)

0.00

2.00

4.00

6.00

8.00

10.00

12.00

14.00

PPer

-Pai

r Per

mis

sibl

e Lo

ad

SPA-(w/o NE,w/IM)-4S ITU-DR ITU-SR IETF-DRIETF-SR SPA-(w/o NE,w/IM)-1S SPA-(w/o NE,w/IM)-2S SPA-(w/o NE,w/IM)-3S

Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. IETF-DR2. ITU-DR3. IETF-SR4. ITU-SR5. SPA-(w/oNE,w/IM)-1S6. SPA-(w/oNE,w/IM)-2S7. SPA-(w/oNE,w/IM)-3S8. SPA-(w/o/NE,w/IM)-4S

Permissible load Key Takeaways: 1. SPA-(w/oNE,w/IM) provides a higher permissible load compred to IETF and ITU models under both direct and split routing.2. For SPA-(w/oNE,w/IM) model, increasing sharing ratio leads to higher permissible load




0 5 10 15 20 25 30 35 40

Input Load (Erlang)

0.00

2.00

4.00

6.00

8.00

10.00

12.00

14.00

16.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. IETF-DR2. ITU-DR3. IETF-SR4. ITU-SR5. SPA-w/(NE,IM)-4S6. SPA-w/(NE,IM)-3S7. SPA-w/(NE,IM)-2S8. SPA-w/(NE,IM)-1S

Permissible load Key Takeaways: 1. SPA-w/(NE,IM) provides a higher permissible load compred to IETF and ITU models under both direct and split routing.2. For SPA-w/(NE,IM) model, increasing sharing ratio leads to lower permissible load3. The significane of sharing ratio of SPA-w/(NE,IM) on permissible load is higher than SPA-(w/oNE,w/IM) model

53Appendix-B: 4 Node Topology Detailed Results “Utilization”

4-node Topology (Fully-meshed Service Configuration)Average Network-Wide Utilization (Physical Resources)

2-Alternate Routing, Class-B Arrivals, IETF,ITU, SPA-Dedicated, SPA-w/o(NE,IM)

0%

20%

40%

60%

80%

100%

120%

0 5 10 15 20 25 30 35 40

Input Load (Erlang)

Phys

ical

Res

ourc

es U

tiliz

atio

n

IETF-DR IETF-SR ITU-DR ITU-SR SPA-DedicatedSPA-w/o(NE,IM )-1S SPA-w/o(NE,IM )-2S SPA-w/o(NE,IM )-3S SPA-w/o(NE,IM )-4S

Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. SPA-w/o(NE,IM)2. ITU-DR3. SPA-Dedicated4. ITU-SR5. IETF-DR6. IETF-SR

Utilization Key Takeaways: Under SPA-w/o(NE,IM) and same input load:1. All sharing ratios provide lower utilization than IETF,ITU, and SPA-Dedicated models.2. SPA-Dedicated provides lower utilization than IETF-DR,ITU-SR, and IETF-SR models.3. Direct Routing (DR) povides lower utlization than Split Routing (SR) for both IETF and ITU models



2-Alternate Routing, Class-B Arrivals, IETF,ITU, SPA-Dedicated, SPA-w/NE,w/oIM

0%

20%

40%

60%

80%

100%

120%

0 5 10 15 20 25 30 35 40

Input Load (Erlang)

Phys

ical

Res

ourc

es U

tiliz

atio

n

IETF-DR IETF-SR ITU-DRITU-SR SPA-Dedicated SPA-(w/ NE,w/oIM)-1 SSPA-(w/ NE,w/oIM)-2S SPA-(w/ NE,w/oIM)-3S SPA-(w/ NE,w/oIM)-4S

Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. SPA-(w/NE,w/oIM)2. ITU-DR3. SPA-Dedicated4. ITU-SR5. IETF-DR6 IETF-SR

Utilization Key Takeaways: Under SPA-(w/NE,w/oIM) and same input load:1. All sharing ratios provide lower utilization than IETF,ITU, and SPA-Dedicated models.2. SPA-Dedicated provides lower utilization than IETF-DR,ITU-SR, and IETF-SR models.3. Direct Routing (DR) povides lower utlization than Split Routing (SR) for both IETF and ITU models



2-Alternate Routing, Class-B Arrivals, IETF,ITU, SPA-Dedicated, SPA-w/oNE,w/IM

40%

50%

60%

70%

80%

90%

100%

110%

0 5 10 15 20 25 30 35 40

Input Load

Phys

ical

Res

ourc

es U

tiliz

atio

n

IETF-DR IETF-SR ITU-DRITU-SR SPA-Dedicated SPA-(w/o NE,w/IM)-1SSPA-(w/o NE,w/IM)-2S SPA-(w/o NE,w/IM)-3S SPA-(w/o NE,w/IM)-4S

Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. ITU-DR2. SPA-Dedicated3. ITU-SR4. IETF-DR5. SPA-(w/oNE,w/IM)6. IETF-SR

Utilization Key Takeaways: Under SPA-(w/oNE,w/IM) and same input load:1. All sharing ratios provide higher utilization than IETF-DR,ITU-SR, ITU-DR, and SPA-Dedicated models.2. SPA-Dedicated provides lower utilization than IETF-DR,ITU-SR, and IETF-SR models.3. Direct Routing (DR) povides lower utlization than Split Routing (SR) for both IETF and ITU models



2-Alternate Routing, Class-B Arrivals, IETF,ITU, SPA-Dedicated, SPA-w/(NE,IM)

0%

20%

40%

60%

80%

100%

120%

0 5 10 15 20 25 30 35 40

Input Load (Erlang)

Phys

ical

Res

ourc

es U

tiliz

atio

n

IETF-DR IETF-SR ITU-DR ITU-SR SPA-DedicatedSPA-(w/ NE,IM)-1S SPA-(w/ NE,IM)-2S SPA-(w/ NE,IM)-3S SPA-(w/ NE,IM)-4S

Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. ITU-DR2. SPA-Dedicated3. ITU-SR4. IETF-DR5. SPA-w/(NE,IM)6. IETF-SR

Utilization Key Takeaways: Under SPA-w/(NE,IM) and same input load:1. All sharing ratios provide higher utilization than IETF-DR, ITU-DR, ITU-SR, and SPA-Dedicated models.2. SPA-Dedicated provides lower utilization than IETF-DR,ITU-SR, and IETF-SR models.3. Direct Routing (DR) povides lower utlization than Split Routing (SR) for both IETF and ITU models

57Appendix-C: 7 Node (2 routes) Topology Detailed Results “Blocking probability”



0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

Blo

ckin

g Pr

obab

ility


Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. ITU-DR2. IETF-DR3.SPA- Dedicated4. ITU-SR5. IETF-SR

Blocking Key Takeaways: Enabling direct routing for both IETF and ITU control plane models leads to lower blocking probability than SPA-Dedicated.




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

Blo

ckin

g Pr

obab

ility


Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. ITU-DR2. IETF-DR3. SPA- w/o(NE,/IM)-1S4. SPA- w/o(NE,/IM)-3S5. SPA- w/o(NE,/IM)-4S6. SPA- w/o(NE,/IM)-2S7. ITU-SR8. IETF-SR

Blocking Key Takeaways: 1. Disabling NE & IM under any sharing ratio leads to higher blocking probability than both IETF-DR and ITU-DR, but lower blocking probability that IETF-SR and ITU-SR2. Increasing sharing ratio leads to higher blocking probability on the SPA-w/o(NE,IM)




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

Blo

ckin

g Pr

obab

ility


Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. ITU-DR2. IETF-DR3. SPA- (w/NE,w/o/IM)-1S4. SPA- (w/NE,w/o/IM)-3S5. SPA- (w/NE,w/o/IM)-2S6. SPA- (w/NE,w/o/IM)-4S7. ITU-SR8. IETF-SR

Blocking Key Takeaways: 1. Enabling NE only under any sharing ratio leads to higher blocking probability than both IETF-DR and ITU-DR, but lower blocking probability that IETF-SR and ITU-SR2. Increasing sharing ratio has no direct effect on blocking probability on the SPA-(w/NE,w/oIM)




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

Blo

ckin

g Pr

obab

ility


Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. SPA- (w/oNE,w//IM)-4S2. SPA- (w/oNE,w//IM)-3S3. SPA- (w/oNE,w//IM)-2S4. SPA- (w/oNE,w//IM)-1S5. ITU-DR6. IETF-DR7. ITU-SR8. IETF-SR

Blocking Key Takeaways: 1. Enabling IM under any sharing ratio leads to lower blocking probability than IETF-DR, ITU-DR, IETF-SR, and ITU-SR2. Under lower input loads, Increasing sharing ratio leads to lower bocking probability on the SPA-(w/oNE,w/IM)




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

Blo

ckin

g Pr

obab

ility


Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. SPA- w/(NE,IM)-1S2. SPA- w/(NE,IM)-2S3. SPA- w/(NE,IM)-3S4. SPA- w/(NE,IM)-4S5. ITU-DR6. IETF-DR7. ITU-SR8. IETF-SR

Blocking Key Takeaways: 1. Enabling both NE & IM under any sharing ratio leads to lower blocking probability than IETF-DR, ITU-DR, IETF-SR, and ITU-SR2. Enabling NE in addition to IM leads to lower blocking probability.

62Appendix-C: 7 Node (2 routes) Topology Detailed Results “Permissible Load”

7-node Topology (Fully-meshed Service Configuration)Average Network-Wide Permissible Load (Physcial Resources)


0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

1.00

2.00

3.00

4.00

5.00

6.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. IETF-DR2. ITU-DR3. SPA-Dedicated4. IETF-SR5. ITU-DRUnder any given input load:1. No significant permissable load advantage of the SPA-Dedicated over both IETF-DR and ITU-DR2. ITU-DR & IETF-DR provides a lower permissable load than (IETF-SR,ITU-SR, and SPA-Dedicated

Permissible load Key Takeaways: 1. Split routing provides higher PL than direct routing for both IETF and ITU control plane models.




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

1.00

2.00

3.00

4.00

5.00

6.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. SPA-w/o(NE,IM)-2S2. SPA-w/o(NE,IM)-4S3. SPA-w/o(NE,IM)-3S4. SPA-w/o(NE,IM)-1S5. IETF-DR6. ITU-DR7. IETF-SR8. ITU-DR

Permissible load Key Takeaways: 1. SPA-w/o(NE,IM) provides lower permissable loadcompred to IETF and ITU models under both direct and split routing.




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

1.00

2.00

3.00

4.00

5.00

6.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. SPA-(w/NE,w/oIM)-4S2. SPA-(w/NE,w/oIM)-2S3. SPA-(w/NE,w/oIM)-3S4. SPA-(w/NE,w/oIM)-1S5. IETF-DR6. ITU-DR7. IETF-SR8. ITU-DR

Permissible load Key Takeaways: 1. SPA-(w/NE,w/oIM) provides a lower permissable load compred to IETF and ITU models under both direct and split routing.2. For SPA-(w/NE,w/oIM) model, increasing sharing ratio leads to lower permissable load




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

2.00

4.00

6.00

8.00

10.00

12.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. IETF-DR2. ITU-DR3. IETF-SR4. ITU-DR5. SPA-(w/oNE,w/IM)-1S6. SPA-(w/oNE,w/IM)-2S7. SPA-(w/oNE,w/IM)-3S8. SPA-(w/o/NE,w/IM)-4S

Permissible load Key Takeaways: 1. SPA-(w/oNE,w/IM) provides a higher permissable load compred to IETF and ITU models under both direct and split routing.2. For SPA-(w/oNE,w/IM) model, increasing sharing ratio leads to higher permissable load3. The significane of sharing ratio of SPA-(w/oNE,w/IM) on permissable load is not major when sharing is above 2 STS.




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

2.00

4.00

6.00

8.00

10.00

12.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. IETF-DR2. ITU-DR3. IETF-SR4. IETF-SR5. SPA-w/(NE,IM)-4S6. SPA-w/(NE,IM)-3S7. SPA-w/(NE,IM)-2S8. SPA-w/(NE,IM)-1S

Load Key Takeaways: 1. SPA-w/(NE,IM) provides a higher permissable load compred to IETF and ITU models under both direct and split routing.2. For SPA-w/(NE,IM) model, increasing sharing ratio leads to lower permissable load3. The significane of sharing ratio of SPA-w/(NE,IM) on permissable load is higher than SPA-(w/oNE,w/IM) model

67Appendix-C: 7 Node (2 routes) Topology Detailed Results “Utilization”



0%

20%

40%

60%

80%

100%

120%

0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

Phys

ical

Res

ourc

es U

tiliz

atio

n


Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. SPA-w/o(NE,IM)3. ITU-DR2. SPA-Dedicated4. IETF-DR5. ITU-SR6. IETF-SR

Utilization Key Takeaways: Under SPA-w/o(NE,IM) and same input load:1. All sharing ratios provide lower utilization than IETF,ITU, and SPA-Dedicated models.2. SPA-Dedicated provides lower utilization than IETF-DR,ITU-SR, and IETF-SR models.3. Direct Routing (DR) povides lower utlization than Split Routing (SR) for both IETF and ITU models




0%

20%

40%

60%

80%

100%

120%

0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

Phys

ical

Res

ourc

es U

tiliz

atio

n


Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. SPA-(w/NE,w/oIM)3. ITU-DR2. SPA-Dedicated4. IETF-DR5. ITU-SR6 IETF-SR

Utilization Key Takeaways: Under SPA-(w/NE,w/oIM) and same input load:1. All sharing ratios provide lower utilization than IETF,ITU, and SPA-Dedicated models.2. SPA-Dedicated provides lower utilization than IETF-DR,ITU-SR, and IETF-SR models.3. Direct Routing (DR) povides lower utlization than Split Routing (SR) for both IETF and ITU models




40%

50%

60%

70%

80%

90%

100%

110%

0 10 20 30 40 50 60 70 80 90

Input Load

Phys

ical

Res

ourc

es U

tiliz

atio

n


Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. ITU-DR2. SPA-Dedicated3. IETF-DR4. ITU-SR5. SPA-(w/oNE,w/IM)6 IETF-SR





0%

20%

40%

60%

80%

100%

120%

0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

Phys

ical

Res

ourc

es U

tiliz

atio

n


Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. ITU-DR2. SPA-Dedicated3. IETF-DR4. ITU-SR5. SPA-w/(NE,IM)6. IETF-SR


71Appendix-D: 7 Node (3 routes) Topology Detailed Results “Blocking probability”



0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Blo

ckin

g Pr

obab

ility


Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. ITU-DR2. IETF-DR3. SPA- Dedicated4. ITU-SR5. IETF-SR

Blocking Key Takeaways: Enabling direct routing for both IETF and ITU control plane models leads to lower blocking probability than SPA-Dedicated.




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Blo

ckin

g Pr

obab

ility


Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. ITU-DR2. IETF-DR3. SPA- w/o(NE,/IM)-1S4. SPA- w/o(NE,/IM)-3S5. SPA- w/o(NE,/IM)-4S6. SPA- w/o(NE,/IM)-2S7. ITU-SR8. IETF-SR

Blocking Key Takeaways: 1. Disabling NE & IM under any sharing ratio leads to higher blocking probability than both IETF-DR and ITU-DR, but lower blocking probability that IETF-SR and ITU-SR2. Increasing sharing ratio leads to higher blocking probability on the SPA-w/o(NE,IM)




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Blo

ckin

g Pr

obab

ility


Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. ITU-DR2. IETF-DR3. SPA- (w/NE,w/o/IM)-3S4. SPA- (w/NE,w/o/IM)-1S5. SPA- (w/NE,w/o/IM)-4S6. SPA- (w/NE,w/o/IM)-2S7. ITU-SR8. IETF-SR

Blocking Key Takeaways: 1. Enabling NE only under any sharing ratio leads to higher blocking probability than both IETF-DR and ITU-DR, but lower blocking probability that IETF-SR and ITU-SR2. Increasing sharing ratio has no direct effect on blocking probability on the SPA-(w/NE,w/oIM)




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Blo

ckin

g Pr

obab

ility


Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. SPA- (w/oNE,w//IM)-4S2. SPA- (w/oNE,w//IM)-3S3. SPA- (w/oNE,w//IM)-2S4. SPA- (w/oNE,w//IM)-1S5. ITU-DR6. IETF-DR7. ITU-SR8. IETF-SR

Blocking Key Takeaways: 1. Enabling IM under any sharing ratio leads to lower blocking probability than IETF-DR, ITU-DR, IETF-SR, and ITU-SR2. Under lower input loads, Increasing sharing ratio leads to lower bocking probability on the SPA-(w/oNE,w/IM)




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Blo

ckin

g Pr

obab

ility


Summary: The following control plane models are listed in ascending order of the Physical Resources blocking probability:1. SPA- w/(NE,/IM)-1S2. SPA- w/(NE,/IM)-2S3. SPA- w/(NE,/IM)-3S4. SPA- w/(NE,/IM)-4S5. ITU-DR6. IETF-DR7. ITU-SR8. IETF-SR

Blocking Key Takeaways: 1. Enabling both NE & IM under any sharing ratio leads to lower blocking probability than IETF-DR, ITU-DR, IETF-SR, and ITU-SR2. Enabling NE in addition to IM leads to lower blocking probability.3. Under lower input loads, Increasing sharing ratio leads to higher bocking probability on the SPA-w/(NE,IM)

76Appendix-D: 7 Node (3 routes) Topology Detailed Results “Permissible Load”

7-node Topology (Fully-meshed Service Configuration)Average Network-Wide Permissible Load (Physcial Resources)


0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

4.50

5.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. IETF-DR2. ITU-DR3. SPA-Dedicated4. IETF-SR5. ITU-DRUnder any given input load:1. No significant permissible load advantage of the SPA-Dedicated over both IETF-DR and ITU-DR2. ITU-DR & IETF-DR provides a higher permissible load than (IETF-SR,ITU-SR, and SPA-Dedicated

Permissible Load Key Takeaway:1. Split routing provides higher permissible load than direct routing for both IETF and ITU control plane models.




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

4.50

5.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. SPA-w/o(NE,IM)-2S2. SPA-w/o(NE,IM)-4S3. SPA-w/o(NE,IM)-3S4. SPA-w/o(NE,IM)-1S5. IETF-DR6. ITU-DR7. IETF-SR8. ITU-DR

Permissible Load Key Takeaway:1. SPA-w/o(NE,IM) provides lower permissible load compred to IETF and ITU models under both direct and split routing.




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

4.50

5.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. SPA-(w/NE,w/oIM)-4S2. SPA-(w/NE,w/oIM)-2S3. SPA-(w/NE,w/oIM)-3S4. SPA-(w/NE,w/oIM)-1S5. IETF-DR6. ITU-DR7. IETF-SR8. ITU-DR

Permissible Load Key Takeaway:1. SPA-(w/NE,w/oIM) provides a lower permissible load compred to IETF and ITU models under both direct and split routing.2. For SPA-(w/NE,w/oIM) model, increasing sharing ratio leads to lower permissible load




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

2.00

4.00

6.00

8.00

10.00

12.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. IETF-DR2. ITU-DR3. IETF-SR4. ITU-DR5. SPA-(w/oNE,w/IM)-1S6. SPA-(w/oNE,w/IM)-2S7. SPA-(w/oNE,w/IM)-3S8. SPA-(w/o/NE,w/IM)-4S

Permissible Load Key Takeaway:1. SPA-(w/oNE,w/IM) provides a higher permissible load compred to IETF and ITU models under both direct and split routing.2. For SPA-(w/oNE,w/IM) model, increasing sharing ratio leads to higher permissible load 3. The significane of sharing ratio of SPA-(w/oNE,w/IM) on permissible load is not major when sharing is above 2 STS.




0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

0.00

2.00

4.00

6.00

8.00

10.00

12.00

Per-

Pair

Perm

issi

ble

Load


Summary: The following control plane models are listed in ascending order of the physical resources permissible load:1. IETF-DR2. ITU-DR3. IETF-SR4. IETF-SR5. SPA-w/(NE,IM)-4S6. SPA-w/(NE,IM)-3S7. SPA-w/(NE,IM)-2S8. SPA-w/(NE,IM)-1S

Permissible Load Key Takeaway:1. SPA-w/(NE,IM) provides a higher permissible load compred to IETF and ITU models under both direct and split routing.2. For SPA-w/(NE,IM) model, increasing sharing ratio leads to lower permissible load 3. The significane of sharing ratio of SPA-w/(NE,IM) on permissible load is higher than SPA-(w/oNE,w/IM) model

81Appendix-D: 7 Node (3 routes) Topology Detailed Results “Utilization”



0%

20%

40%

60%

80%

100%

120%

0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

Phys

ical

Res

ourc

es U

tiliz

atio

n


Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. SPA-w/o(NE,IM)3. ITU-DR2. SPA-Dedicated4. IETF-DR5. ITU-SR6. IETF-SR

Utilization Key Takeaways: Under SPA-w/o(NE,IM) and same input load:1. All sharing ratios provide lower utilization than IETF,ITU, and SO-Dedicated models.2. SPA-Dedicated provides lower utilization than IETF-DR,ITU-SR, and IETF-SR models.3. Direct Routing (DR) povides lower utlization than Split Routing (SR) for both IETF and ITU models




0%

20%

40%

60%

80%

100%

120%

0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

Phys

ical

Res

ourc

es U

tiliz

atio

n


Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. SPA-(w/NE,w/oIM)3. ITU-DR2. SPA-Dedicated4. IETF-DR5. ITU-SR6 IETF-SR

Utilization Key Takeaways: Under SPA-(w/NE,w/oIM) and same input load:1. All sharing ratios provide lower utilization than IETF,ITU, and SO-Dedicated models.2. SPA-Dedicated provides lower utilization than IETF-DR,ITU-SR, and IETF-SR models.3. Direct Routing (DR) povides lower utlization than Split Routing (SR) for both IETF and ITU models




40%

50%

60%

70%

80%

90%

100%

110%

0 10 20 30 40 50 60 70 80 90

Input Load

Phys

ical

Res

ourc

es U

tiliz

atio

n


Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. ITU-DR2. SPA-Dedicated3. IETF-DR4. ITU-SR5. SPA-(w/oNE,w/IM)6 IETF-SR





0%

20%

40%

60%

80%

100%

120%

0 10 20 30 40 50 60 70 80 90

Input Load (Erlang)

Phys

ical

Res

ourc

es U

tiliz

atio

n


Summary: Based on the physical resources utilization, the following control plane models are listed in ascending order based on the physical resources utlization under the same input load1. ITU-DR2. SPA-Dedicated3. IETF-DR4. ITU-SR5. SPA-w/(NE,IM)6. IETF-SR


Documents

PhD Dissertation Defense - ITTC€¦ · PhD Dissertation Defense Service Profile-Aware Control Plane: A Multi-Instance Fixed Point Approximation within A Multi-Granularity VPN Loss