GRID.IT GRID.IT Programma Nazionale della Ricerca 2001-03 Bando FI - - PDF document

Progetto

“Piattaforme abilitanti per griglie computazionali a elevate prestazioni

• rientate a organizzazioni virtuali scalabili”

GRI D.I T

(Resp.: Prof. Marco Vanneschi, UniPi/ CNR)

CHECK POI NT – 1° anno

WP1 – GRI D ORI ENTED OPTI CAL SWI TCHI NG PARADI GMS

• Resp. Piero Castoldi

Pisa, 6-8 Ottobre 2003

Programma Nazionale della Ricerca 2001-03 Bando FI RB

Progetto Strategico “Tecnologie Abilitanti per la Società della Conoscenza” Progetto obiettivo “Reti e Netputing”

GRID.IT GRID.IT

2

GRID.IT GRID.IT

Relazione con l’UdR CNIT

• Resp. UdR CNIT: Prof. Giancarlo Prati

Articolazione su due workpackages

• WP 1 – Grid oriented optical switching

paradigms (Resp. Piero Castoldi)

• WP 2 - High-performance photonic test-bed

(Resp. Stefano Giordano)

3

GRID.IT GRID.IT

Specifications on interaction among areas

WP10 WP11 WP12 WP13 WP8 WP9 WP7 WP4 WP6

Area 1: Middleware and Programming Tools Area 4: Applications

WP14 WP3 WP5 WP1 WP2

A1.1, A1.2, A1.4 A1.1, A1.2, A1.3, A1.4, A1.5

Area 2: Photonic testbed Area 3: Grid deployment

4

GRID.IT GRID.IT

Istituzioni che partecipano al al WP1

• Laboratorio Nazionale di Reti Fotoniche, Pisa

Know-how: all 5 major areas of optical networks and photonic technologies (routing and switching, trasmission, amplification, systems)

• UdR CNIT at

Università Politecnica delle Marche

Know-how: enabling technologies and OXC WXC architectures

Università di Bologna

Know-how: architectures for optical packet switching networks

Università di Trento

Know-how: traffic models and resource allocation

• University of Texas @ Dallas (UTD)

Know-how: control plane and reliability for optical networks

5

GRID.IT GRID.IT

WP1 breakdown

WP 1 - Grid oriented optical switching paradigms (Resp. P. Castoldi)

[70% of the total CNIT UdR funding]

• Activity 1 – Connections, topologies and network service models

(Resp. R. Battiti) [Lab PI , TN]

• Activity 2 – Grid computing on state-of-the-art optical networks

(Resp. P. Castoldi) [Lab PI , TN, UTD]

• Activity 3 – Migration scenarios to intelligent flexible optical networks (Resp. F. Callegati)

[Lab PI , BO, TN, UTD]

• Activity 4 – Control plane and network emulation for optical packet switching networks

(Resp. A. Fumagalli) [Lab PI , TN, UTD]

• Activity 5 – Enabling technologies for optical switching networks

(Resp. G. Cancellieri) [Lab PI , BO, AN]

1° year 2° year 1° year 2° year 3° year 1° year 2° year 2° year 3° year 1° year 2° year 8% 15% 15% 15% 17%

Technical Board member: Dr. Luca Valcarenghi (assistant professor).

6

GRID.IT GRID.IT

Optical Grid Networking Roadmap

f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2)

Emitter (E)/ Collector (C) Virtual Processor (VP) Distributed Virtual Shared Memory (DVSM) connection

7

GRID.IT GRID.IT

Details about activity 1

8% Activity 1 – Connections, topologies and network service models

(Resp. R. Battiti) [Lab PI, TN, UTD]

1° year 2° year

MILESTONES and DELIVERABLES (1-18 months)

Traffic models to represent computer generated loads

[technical report]

Resource Allocation Architecture (RAA) to map network

logical topology requests from GRID to the optical network and virtual topology definition (VTD) based on given traffic patterns [technical report and software tool]

Added value for GRI D.I T project: Use of traffic models for performance evaluation / capacity planning (input to other activities) Efficient traffic engineering rules for state of the art grid-oriented networks

f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2) f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2)

f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2) f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2)

8

GRID.IT GRID.IT

Achieved Results in Traffic Models (1)

• Development of a WinPCap-based tool for traffic analysis (for

comparisons between models and actual traffic). Ddum p is a useful tool for traffic capture and analysis and it provides several functionalities: Compatibility with libpcap and interoperability with Ethereal Tree and database packet representation Packet arrival / interarrival time distribution graphs Distribution matching with traffic models (Poisson, binomial) Possibility of packet header modification / trace file editing Data export to libpcap / ns-2-compatible files Traffic generation Network status evaluation (ping, traceroute) Written in MS Visual C+ + w/ WinPcap libraries

9

GRID.IT GRID.IT

Achieved Results in Traffic Models (2)

An open-source tool for traffic measurement and analysis, called Ddump, available online

http://dit.unitn.it/∼ddump

10

GRID.IT GRID.IT

Achieved Results in Traffic Models (3)

Validation of models for complete TCP characterization (closed form) by simulation

1000 2000 3000 4000 5000 6000 7000 8000 9000 0,001 0,005 0,01 0,5 1 Sim Model 2000 4000 6000 8000 10000 1 5 10 15 20 30 Sim Model

Average throughput using NS-2 simulator (“Sim”) and the proposed model (“Model”) for different numbers of TCP sources and a fixed bottleneck delay.

Σ F2 F2 F2 Single TCP source λ1 µ1 λ2 µ2 λN µN λ µ B Tw,1 Tw,2 Tw,N Bottleneck

Average throughput (Mbps) for different ACK delay values using NS-2 simulator (“Sim”) and the proposed model (“Model”). 10 TCP sources transmitting at 1Mbps are considered. Single source m odel: M/ P/ 1/ cwnd queue, i.e. Poisson arrivals (M), TCP parametrized service distribution (P), 1 server, cwnd= 64 packets, outgoing link 1 Mbps, packets of 1024 bits Bottleneck ( router) m odel: modeled as M/ M/ 1/ B, i.e. with Poisson arrivals and service time, 1 server, buffer capacity of B= 128 packets, outgoing link of 10 Mbps

11

GRID.IT GRID.IT

Achieved results on RAA and VTD (1) 1) FID (x), First Improve Dynamic load balancing 2) LFID, Lazy First-Improve Dynamic Load balancing

re-route a single LSP according to FID only if the new one cannot be accommodated 1. Route the LSPs using some routing scheme (CBR) 2. Repeat for MAXITER iterations: a. Find the most congested edges (MCE) in the network (residual capacity less than x) b. For each most congested edge:

• try to re-route only one of the LSPs on the MCE over

another edge

• IF congestion is reduced, accept move, and ask the

ingress LSR to re-route this LSP

12

GRID.IT GRID.IT

Achieved Results in RAA and VTD (2)

Comparison with existing preventive RAA algorithms (MHA, i.e. minimum hop routing and MIRA, i.e. minimum interference routing).

Number of rejected LSPs vs. network load

Parameters:

• Arrival rate with rate λ
• Duration of the

transactions 1/µ

• Threshold for FID
• peration 0.01
• 12 units of capacity in

the light link

• 48 units of capacity in

the dark links

• LSP bandwidth demand

in between 0.01 and 0.04 units of capacity

13

GRID.IT GRID.IT

Publications on activity 1 of WP1

1.

• D. Agrawal, F. Granelli, A Queuing Model for Steady-State Behaviour of

TCP in Performance Evaluation of Telecommunication Networks, Softcom 2003, Oct. 2003. 2.

• M. Brunato, R. Battiti, E. Salvadori, Load Balancing in WDM Networks

through Adaptive Routing Table Changes. In Networking, number 2345 in Lecture Notes in Computer Science, pages 289–301, Pisa - Italy, May

• 2002. Springer Verlag.

3.

• M. Brunato, R. Battiti, E. Salvadori. Dynamic Load Balancing in WDM
• Networks. Optical Networks Magazine, September 2003. In press.

4.

• E. Salvadori, R. Battiti. A Load Balancing Scheme for Congestion Control

in MPLS Networks. Accepted in IEEE Symposium on Computers and Communications – ISCC 2003, Antalya - Turkey. 5.

• E. Salvadori, R. Battiti, F. Ardito. Lazy Rerouting for MPLS Traffic
• Engineering. Submitted to IEEE Globecom 2003.

+ 4 DIT Technical Reports

14

GRID.IT GRID.IT

Details about activity 2

1° year 2° year 15% Activity 2 – Grid computing on state-of-the-art optical networks

(Resp. P. Castoldi) [Lab PI , TN, UTD] MILESTONE and DELIVERABLES (1-12 months)

Static and dynamic routing and wavelength assignment (RWA) of

• ptical connections [technical report and software tools]

MILESTONE and DELIVERABLES (6-18 months)

Reliability schemes exploiting traffic granularity (GMPLS) for

protection and restoration [technical report and laboratory testbed]

GMPLS based control plane for dynamic distributed resource

allocation [technical report and laboratory testbed]

15

GRID.IT GRID.IT

Routing and Wavelength Assignment (RWA)

What has been promised:

• Blocking probability of static and dynamic RWA schemes
• Assessment of benefits of wavelength selection and/or

wavelength conversion

• Reduction of complexity by sparse or limited wavelength

selection and conversion

Advantages to GRID.IT

• Guidelines for optical network testbed implementations

based on a circuit switching paradigm

• Complexity requirements for the nodes

f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2) f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2)

16

GRID.IT GRID.IT

RWA - Considered Node architectures (1)

• No wavelength selection
• No wavelength conversion
• Wavelength selection
• No wavelength conversion

Present situation

17

GRID.IT GRID.IT

RWA – Considered Node architectures (2)

• No wavelength selection
• Wavelength conversion
• Wavelength selection
• Wavelength conversion

Best performance Most expensive

18

GRID.IT GRID.IT

Simulation results – Blocking probability

WS NOWC WS NOWC NOWS WC NOWS WC

19

GRID.IT GRID.IT

Simulation results – Average link usage

NOWS NOWS NOWC NOWC WS WS NOWC NOWC

20

GRID.IT GRID.IT

Sparse wavelength selection

WS property

• f every node

Sparse WS is

not beneficial

21

GRID.IT GRID.IT

Sparse and limited wavelength conversion

WC property of the

whole network

Just few converters in

some nodes improve the performance

22

GRID.IT GRID.IT

Work on Resilience for grid optical networks

What has been promised:

• Integrated (applications + network)

survivability schemes

• Preplanned and on-the-fly restoration

schemes at the network layer

• Exploitation of traffic granularity at the

network layer

Advantages to GRID.IT

• Pushing the limits of resilience

efficiency in optical networks (very close to that of electrical networks)

• Extended robustness for networks

supporting grid computing

f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2) f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2)

f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2) f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2) f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2) f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2)

23

GRID.IT GRID.IT

Current Approach Characteristics and Study Objective

Current failover schemes

• Advantages

Network layer independent Flexible (e.g., degree of failover dependent on application)

• Drawbacks

Application dependent User driven Need for TCP synchronization Slow reaction to failures Not scalable (e.g., CPU and storage)

Study objective

• Exploiting the interaction between resilient approaches based on

task and data replication/migration with network layer resilient schemes

24

GRID.IT GRID.IT

Grid Networking Resilience Issues

f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2)

Emitter (E)/ Collector (C) Virtual Processor (VP) Distributed Virtual Shared Memory (DVSM) connection

Motivations

Guaranteeing the successful calculation of function f(x,d) in the presence or absence

• f grid infrastructure

failures Combining current failover schemes implemented in the grid infrastructure

25

GRID.IT GRID.IT

Proposed Approach

• Combining network layer connection rerouting with

task/data replica placement

• ILP formulation objective
• maximizing the number of connections restored after

failure

1 3 5 B

• riginal task/data

task/data replica 2 4 B G H A D 2 4 B G H A D 1 5 3 2 4 B G A D 5 3 H A B D G H

26

GRID.IT GRID.IT

Results on integrated resilience schemes

• Simulation scenario
• 10 node and 22 link network, 8

wavelengths per link, single link failure

• 200 randomly generated connection

matrices

• Bidirectional connection generation
• Unidirectional connection rerouting
• Expected restoration blocking probability
• average ratio between number of

blocked connections and failed connections

• Expected fraction of replica location per

failed (s,d) pair

• average ratio between number of utilized

replica locations (original destination excluded) and allowed number of replica locations

• Replica utilization lowers restoration

blocking probability

• Replica location utilization decreases for

high throughput

27

GRID.IT GRID.IT

Conclusions on integrated resilience schemes

Exploited synergy between network layer failover

schemes (e.g., connection rerouting) and upper layer (e.g., transport, application) failover schemes (e.g., task/data replication/migration)

Task/data replication/migration improves restoration

blocking probability

Low efficiency of task/data replica utilization for high

throughput

Need to define a policy for replica staging 28

GRID.IT GRID.IT

Network Restoration Schemes

The The I ntegrated I ntegrated Multi Multi layer approach tries to layer approach tries to exploit exploit GLSPs finer GLSPs finer granularity to granularity to reduce reduce blocking probability blocking probability

S S D D

Working Working path from path from S S to to D D

Spare capacity Spare capacity 1.2 1.2 0.6 0.6 0.4 0.4 1.2 1.2 1.2 1.2

• Single layer
• 1. Heuristic choice
• 2. Deterministic choice
• Exhaustive
• Iterative
• Integrated Multi layer
• 1. Heuristic choice
• 2. Grooming capable node

29

GRID.IT GRID.IT

Results: Heuristic choice

Single Single-

• Multi

Multi layer comparison layer comparison

Grooming Nodes NO Grooming Nodes

• Good improvement of the Restoration

Blocking Probability;

• Limited necessary grooming capability;
• Lower blocking probability also without

grooming capable node.

• Toroidal 4x4 network

topology

• 32 wavelengths link

capacity

• Three preplanned

recovery paths

d s d s 30

GRID.IT GRID.IT

Results: Deterministic choice

Heuristic Heuristic-

• Deterministic choice comparison

Deterministic choice comparison

• Significant

improvement of Heuristic choice.

• Iterative scheme

results equivalent to the Exhaustive

• ne.

31

GRID.IT GRID.IT

Summary on network restoration schemes

Choice policies

• The Multi layer Heuristic choice offers a good improvement of the

blocking probability if grooming capable node are available;

• The Single layer Deterministic choice offers a significant improvement of

the Single layer Heuristic choice;

• The more scalable Iterative Single layer Deterministic choice results

equivalent to the Exhaustive implementation in the simulated scenario. Signaling scheme

• The Source-Destination oriented signaling scheme results is scalable then

the GLSPs oriented;

• The GLSPs oriented signaling scheme is potentially faster due to the

independence of the restoration signaling.

32

GRID.IT GRID.IT

Control plane and data plane of an optical network

• Two planes: data plane and IP based control plane. Worst case assumption: data plane in the
• ptical domain has not the ability of reading the data flowing in it.
• Adjacency discovery cannot occur in the data plane, could occur in control plane if physical

topology of the control plane were the same of the data plane.

• Consequence: the physical topology adjacencies of the data plane with full description of the optical

properties of the node ARE manually loaded into a management element (e.g. a selected PC of the control plane). Problem: how do we describe it? Database …

signaling signaling data data CLIENT CLIENT Optical transport network IP based Optical control plane OXC OXC OXC OXC

33

GRID.IT GRID.IT

• To obtain the best possible performance from the 16 4x1 switches

it was studied a control circuit able to contemporary drive all the switches.

• Input of the control circuit is 1 byte sent from the TTL port of a PC
• Realization:

(1) Truth Table (2) Karnaugh Maps (3) Physical Layout

Development of an OXC prototype and control logics

34

GRID.IT GRID.IT

(4) Implementation (5) Testing (6) Packaging in a box

OXC Physical Realization

35

GRID.IT GRID.IT

Publications on activity 2 of WP1

1.

• L. Valcarenghi and P. Castoldi, “On the Advantages of Integrating Service Migration and Connection Rerouting for

Grid Computing Fault Tolerance”, submitted to ICC2004. 2.

• M. Rameshkumar, P. Castoldi, R. Gangopadhyay, "Virtual wavelength path algorithm for optimal placement of

wavelength converters in WDM cross-connect networks", Proc. of Photonics 2002 Conf., Mumbai (India), 16-18 Dec. 2002. 3.

• P. Castoldi, F. Cugini,"Efficient Design of Differentiated Reliability (DiR) Optical Ring networks with low demand
• utage", Proc. of Photonics 2002 Conf., Mumbai (India), 16-18 Dec. 2002.

4.

• R. Gangopadhyay, C. Kumar, R. Ramana, M. Rameshkumar, P. Castoldi, "Dynamic routing in WDM networks with

bandwidth guaranteed path protection against single link failure", Proc. of Photonics 2002 Conf., Mumbai (India), 16- 18 Dec. 2002. 5.

• M. Rameshkumar, P. Castoldi, R. Gangopadhyay, "The virtual wavelength path approach for optimal placement of

wavelength converters in WDM networks", submitted to SPIE/Kluwer Optical Networks Magazine, April 2003 6.

• N. Andriolli, L. Valcarenghi, P. Castoldi, "Impact of node architecture on the performance of wavelength routed

networks", submitted to JSAC, August 2003. 7.

• S. Gorai, A.Dash, R. Gangopadhyay, P. Castoldi, “Design Algorithm for WDM Network Planning with Joint Wavelength

and waveband routing”, submitted to ICC2004. 8.

• P. Castoldi, F. Cugini, A. Giorgetti, L. Valcarenghi, "Exploiting Stochastic Granularity for Integrated Multilayer

Restoration of GMPLS networks", Proceedings of Photonics in Switching 2003 (PS2003), Paris, 28 September- 2 October 2003. 9.

• F. Cugini, L. Valcarenghi, P. Castoldi, G. Ippoliti, "Experimental Demonstration of Low-Cost 1:1 Optical Span

Protection ", Globecom 2003 - Workshop #4 “Protection and Restoration: from SONET/SDH to Next-Generation Networks”, San Francisco, December 1-5, 2003. 36

GRID.IT GRID.IT

Details about activity 3

15% Activity 3 – Migration scenarios to intelligent flexible optical networks

(Resp. F. Callegati) [Lab PI, BO, TN]

1° year 2° year

MILESTONE and DELIVERABLES [1-12 months]

Scaling issues of optical networks granularity from the flow, to the

burst up to the packet [technical report] MILESTONE and DELIVERABLES [12-24 months]

Optical Packet Switching (OPS) for grid computing: packet formats,

control plane, and OPS applicability limitations [technical report and design tool]

Fast bandwidth reservation scheme for both data transfers of various

size and real-time packets [technical report and design tool]

37

GRID.IT GRID.IT

What has been promised?

Format of the optical packets that best

matches the requirements of data traffic resulting from grid computing applications (exploit results from activity 1)

Function and requirements of the OPS

network control plane tailored to the support

• f connections for grid computing applications

Limits and opportunities of the optical

switching technology that may impair or favour the application of OPS to the grid computing communication scenario

f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2) f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2)

38

GRID.IT GRID.IT

Assumptions

DWDM links Wavelength conversion available Routing is wavelength independent

• All wavelengths of the same fiber are routing equivalent

Congestion resolution in nodes by

• Multiplexing in the wavelength domain
• Multiplexing in the time domain (fiber delay lines)

For the time being simple and general traffic models

• On-off traffic generate as output from a M/M/1 queue

39

GRID.IT GRID.IT

Packet format

Unslotted variable length (UVL) Slotted – fixed length (FLP) Slotted – variable length (SVLP)

T

SVLP FLP UVL

net load = ρ real load = ρS < ρF real load = ρF > ρ

40

GRID.IT GRID.IT

Wavelength and Delay Selection Problem

The forwarding algorithm determines:

the output fiber and the output wavelength if wavelengths busy

packet delayed in FDL buffer or packet dropped, because the required delay is not available

Wavelength and delay selection (WDS) are correlated

minimize the voids maximize the wavelength utilization d

GAP

t0 t0+d t0+2d t0+4d t0+3d

λ1 λ2 λ3 λ4

d

Performance

41

GRID.IT GRID.IT

UVL – packet loss probability

1.0e-006 1.0e-005 1.0e-004 1.0e-003 1.0e-002 1.0e-001 1.0e+000 0.5 1 1.5 2 2.5 3 3.5 4

RND RR RNE RNF MINL MING

Packet Loss Probability d (normalized to the average packet length)

Input Load per wavelength = 0.8 Random traffic N = 4 w = 16 B = 8

42

GRID.IT GRID.IT

FLP vs. SVLP

SVLP

Per train processing and forwarding Worse queuing performance ☺ Lower processing effort in the nodes ☺ Less Overhead

FLP

Slot by slot processing and forwarding ☺ Better queuing performance Higher processing effort in the nodes More Overhead

T

SVLP FLP

43

GRID.IT GRID.IT

Summary on OPS for grid computing

Analysis of different optical packet format

SVLP appears the best trade off between performance and complexity

Dimensioning WDM optical buffers for asynchronous, variable-

length packets

• ptimal value of the buffer delay unit

efficient wavelength allocation algorithm

Joint exploitation of wavelength and time domains

• ptimization leads to good performance with few fiber delay lines

Connectionless scenario

MING is the best choice because it focuses on the minimization of the excess length

Connection-oriented scenario

dynamic wavelength allocation exploits packet correlation

44

GRID.IT GRID.IT

Publications on activity 3 of WP1

1. Franco Callegati, Walter Cerroni, Carla Raffaelli, Paolo Zaffoni: “DWDM for QoS Management in Optical Packet Switches”. QoS-IP 2003: 447-459 2002 2. Franco Callegati, Walter Cerroni, Carla Raffaelli, Paolo Zaffoni: “Dynamic DWDM Exploitation in Connection-Oriented Optical Packet Switches”. ONDM 2002: 151-166

45

GRID.IT GRID.IT

Details about activity 1.5

1° year 2° year 17% Activity 1.5 – Enabling technologies for optical packet switching networks

(Resp. G. Cancellieri) [Lab PI, BO, AN]

3° year

MILESTONE and DELIVERABLES [1-12 months]

Preliminar results on self-routing and automatically switched optical nodes for grid

computing based on Wavelength Routing Switches (WRS) [technical report] MILESTONE and DELIVERABLES [12-24 months]

Final results on self-routing and automatically switched optical nodes for grid

computing based on Wavelength Routing Switches (WRS) [technical report]

Qualifying technologies for optical networks, such as Raman Amplifiers (RA) and

Photonic Christal Fiber (PCF) [technical report] MILESTONE and DELIVERABLES [24-36 months]

Testbed based on automatically switched optical nodes and qualifying optical

tecnologies [laboratory testbed]

46

GRID.IT GRID.IT

What has been promised?

Efficient implementations for WRS to be deployed in

all-optical networks

Preliminary studies about deployment of Raman

amplification techniques in all-optical grid networks

f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2) f(x,d) f(x1,d) f(x2,d1) f(xn,dn) (d1,dn) (d2)

47

GRID.IT GRID.IT

A wavelength recognizing switch (WRS)

• Port I: λr for signaling and λs for data
• port C: is the input for the control wavelength λc
• if λc matches λr, λs is routed towards port T
• if λc does not matches λr, λs is routed towards port P

48

GRID.IT GRID.IT

Progetto architetture 2x2 WRS: cross state

I P T C I P T C IN0 λc IN1 OUT0 OUT1 λc λr0 λs0 λs1 λr1 λr0 = λc: il segnale λs0 deflette verso l’uscita OUT1 λr1 ≠ λc: il segnale λs1 viene inviato verso l’uscita OUT0

49

GRID.IT GRID.IT

OUT0 OUT1 I P T C I P T C IN0 λc IN1 λc λr0 λs0 λs1 λr1 λr0 ≠ λc: il segnale λs0 è inviato verso l’uscita OUT0 λr1 = λc: il segnale λs1 viene deflesso verso l’uscita OUT1

Progetto architetture 2x2 WRS: cross state

50

GRID.IT GRID.IT

Progetto configurazione autoinstradante (1)

Esempio di connessione: input 001→output 100 Lo stato degli switch sull’ingresso di controllo determina verso quale uscita è instradato l’ingresso

51

GRID.IT GRID.IT

00 01 11 10 00 01 11 10 00

01

11 10

000 001 010 011 100 101 110 111 000 001 010 011 100 101 110 111

Esempio di connessione: input 001→output 100 Self-routing bit a bit: ciascun bit dell’indirizzo della porta di uscita comanda la commutazione del singolo elemento 2×2, secondo lo schema bit 0=output “alto”, bit 1=output “basso”.

Progetto configurazione autoinstradante (2)

52

GRID.IT GRID.IT

Il problema del crosstalk

• interazione indesiderata tra i segnali ottici all’interno di uno switch

2×2;

• il segnale commutato verso OUT0 si accoppia anche sulla porta di

uscita OUT1

Effetti:

• degradazione del segnale;
• limitato numero di stadi in cascata per la rete di

commutazione

Crosstalk:

53

GRID.IT GRID.IT

• un solo ingresso attivo per ogni modulo elementare di commutazione;
• implementazione di tecniche RTDM (Reconfiguration with Time Division

Multiplexing). Riduce il throughput!

• implementazione di tecniche a divisione spaziale (matrici di commutazione

replicate e sovrapposte) e definizione di algoritmi di routing interno (multi- colorabilità)

Soluzione per la riduzione del cross-talk

1 2 3 1 2 3

OG λi λi

Per evitare la contemporanea presenza di 2 segnali agli ingressi di un elemento di commutazione, si provvede ad un disaccoppiamento iniziale, inviando tali segnali verso matrici distinte, fra quelle disponibili in diversità verticale di spazio. OG = optical gate, S = 1 x 2 splitter, C = 1 x 2 combiner

S C

54

GRID.IT GRID.IT

1,E-07 1,E-05 1,E-03 1,E-01 4 8 12 16 20 24 28 32 N Permutazioni CF [%]

Senza algoritmi di routing Con bicolorabilità (diversità spaziale di ordine 2)

Risultati sulla bi-colorabilità

55

GRID.IT GRID.IT

Automated Raman Amplifiers characterization

• Raman gain coefficient
• Fiber spectral loss
• Double Rayleigh Sacattering Noise
• Noise Figure
• Spectral gain
• Pump Relative Intensity Noise

ESA OSA AOM Power Meter Photo Diode Pump TLS WFG PC

56

GRID.IT GRID.IT

Publications on activity 5 of WP1

• A.Borella, G.Cancellieri, “Optical burst switching in a WRS all-optical

architecture”, Proc. of 7th European Conference on Networks & Optical Communications, Darmstadt, June 2002, pp. 378-385.

• G. Sacchi, S. Sugliani, G. Bolognini, S. Faralli, F. Di Pasquale “Experimental analysis
• f gain clamping techniques for lumped Raman amplifiers”, ECOC 2003, Rimini, Italy,

paper Tu3.2.4.

• S. Sugliani, G. Sacchi, G. Bolognini, S. Faralli, F. Di Pasquale ,”Effective suppression
• f penalties induced by parametric nonlinear interaction in distributed Raman amplifiers

based on NZ-S fibers”, to be published in IEEE PTL.

57

GRID.IT GRID.IT

Considerazioni generali, problemi, stato avanzamento

• Colmare il gap culturale tra il know how dell’UdR CNIT e il mondo dei

“gridaioli” stricty speaking, still not filled yet .. but positive to reach the handshake within the 2° year.

• Parlare il giusto linguaggio con gli interlocutori dell’area 1 (middleware

& programming tools) e dell’area 3 (grid deployment)

• Fare massa critica dei risultati ottenuti dalle varie “working unit”,

workshop di UdR molto utili a questo scopo.

• Stato avanzamento lavoro sul primo anno: 90%.
• To be done: rifinitura dei deliverable già consegnati, impostazione

dettagliata dell’attività del 2° anno.

58

GRID.IT GRID.IT

ACK to all the people who have contributed!

D'Errico Antonio Lab Naz. Reti Fotoniche CNIT Cococo A1.5 Poli Federica Lab Naz. Reti Fotoniche CNIT Cococo A1.5 Fumagalli Andrea

• UNIV. OF TEXAS @ DALLAS

USA A1.2,A1.3,A1.4 Cerutti Isabella

• UNIV. OF TEXAS @ DALLAS

USA A1.2,A1.3,A1.4 Pitchumani Sudhakar

• UNIV. OF TEXAS @ DALLAS

USA A1.2,A1.3,A1.4 TOTALE Personale A2 Lab Naz. Reti Fotoniche CNIT (Castoldi) + Texas Cugini Filippo Lab Naz. Reti Fotoniche CNIT Ricercatore CNIT A1.2,A1.3,A1.4 Sugliani Simone Lab Naz. Reti Fotoniche CNIT Ricercatore CNIT A1.5 TOTALE Giovani Ric. Lab Naz. Reti Fotoniche CNIT (Castoldi) TOTALE PERSONALE Lab Naz. Reti Fotoniche CNIT (Castoldi) + Texas Russo Franco Lab Naz. Reti Fotoniche CNIT

• Prof. Ordinario

WP2 Giordano Stefano Lab Naz. Reti Fotoniche CNIT

• Prof. Associato

WP2 Adami Davide Lab Naz. Reti Fotoniche CNIT Ricercatore CNIT WP2 Pagano Michele Lab Naz. Reti Fotoniche CNIT Ricercatore WP2 TOTALE Personale A1 Lab Naz. Reti Fotoniche CNIT (Giordano) Repeti Matteo Lab Naz. Reti Fotoniche CNIT Cococo WP2 Orlandini Federico Lab Naz. Reti Fotoniche CNIT Cococo WP2 TOTALE Personale A2 Lab Naz. Reti Fotoniche CNIT (Giordano) TOTALE PERSONALE Lab Naz. Reti Fotoniche CNIT (Giordano) Corazza Giorgio UdR CNIT BO

• Prof. Ordinario

A1.3, A1.5 Raffaelli Carla UdR CNIT BO Ricercatore A1.3, A1.5 Callegati Franco UdR CNIT BO

• Prof. Associato

A1.3, A1.5 Zaffoni Paolo UdR CNIT BO Dottorando A1.3, A1.5 TOTALE Personale A1 UdR CNIT BO (Callegati) Cerroni Walter UdR CNIT BO Ricercatore CNIT A1.3, A1.5 TOTALE Giovani Ric. UdR CNIT BO (Callegati) TOTALE PERSONALE UdR CNIT BO (Callegati) Battiti Roberto UdR CNIT TN

• Prof. Ordinario

A1.1,A1.2,A1.3,A1.4 Granelli Fabrizio UdR CNIT TN Ricercatore A1.1,A1.2,A1.3,A1.4 Salvadori Elio UdR CNIT TN A1.1,A1.2,A1.3,A1.4 Sandro Pera UdR CNIT TN A1.1,A1.2,A1.3,A1.4 TOTALE PERSONALE UdR CNIT TN (Battiti) Cancellieri Giovanni UdR CNIT AN

• Prof. Ordinario

A1.5 Pierleoni Paola UdR CNIT AN Ricercatore A1.5 Gambi Ennio UdR CNIT AN Ricercatore A1.5 TOTALE PERSONALE UdR CNIT AN (Cancellieri)

TOTALE GENERALE PERSONALE 1° ANNO