Intuitive Critical thinking: The Polder Meta Figuring Inititiative.


80 views
Uploaded on:
Description
Intelligent Critical thinking: The Polder Meta Registering Inititiative Subside Sloot Computational Science College of Amsterdam, The Netherlands Ariadne's Red-Rope From PSE to Virtual Research facility and Inspiration Engineering Foundation Work Level: Progressive Booking
Transcripts
Slide 1

Intuitive Problem Solving: The Polder Meta Computing Inititiative Peter Sloot Computational Science University of Amsterdam, The Netherlands

Slide 2

Ariadne’s Red-Rope From PSE to Virtual Laboratory and Motivation Architecture Infrastructure Job Level: Hierarchical Scheduling Resource administration: Task-relocation Interaction && Case usage Interactive Algorithms

Slide 3

Virtual Laboratory Environment Advanced Scientific Domains Computational Physics System Engineering Computational Bio-pharmaceutical Local User Local User Virtual Simulation & Exploration Environment (ViSE) Communication & coordinated effort (ComCol) Virtual-lab Information Management for Cooperation (VIMCO) Physical mechanical assembly Distributed Computing & Gigabit Local Area Network ViSE Net Client App. Client MRI/CT Internet 2 Wide Area Network

Slide 4

Interactive Computing: Why? Objective: From Data, through Information to Knowledge Complexity: Huge information sets, complex procedures Approach: Parametric investigation and affectability examinations: Combine crude (tactile) information with recreation Person on top of it: Sensory communication Intelligent easy routes.:

Slide 5

Intro: Case study from biomedicine...

Slide 6

In Vitro In Vivo In Silico Changing the Paradigm

Slide 7

In Vitro In Vivo In Silico Changing the Paradigm

Slide 8

In Vitro In Vivo In Silico Changing the Paradigm

Slide 9

Diagnosis & Planning Treatment Observation Current Situation

Slide 10

Fast, High-throughput Low Latency Internet High Performance Super Computing New Possibilities in the VL Time and Space Independence 3D Information Simulation based arranging Surgeon ‘in the loop’

Slide 11

Experimental set-up

Slide 12

Architecture

Slide 13

Cave Origine 2000 9 10 11 12 13 14 8 15 7 16 6 17 5 18 4 19 ATM 3 20 2 1 0 23 22 21 GRAPE1 GRAPE0 Architecture Continued: Hybrid framework Host: The DAS 24 hub parallel group in a 200 hub wide range machine 200 MHz Pentium Pro Myrinet 150MB/s ATM wide-territory interconnect between bunches

Slide 14

Immersive Environments

Slide 15

3D Information and Interaction

Slide 16

Problem: Curse of elements: Static undertaking burden Dynamic assignment load Static errand allotment Predictable reallocation Dynamical reallocation Static asset load Dynamic asset load

Slide 17

Solution To Curse Performance of a parallel program typically managed by slowest assignment Task asset prerequisites and accessible assets both differ progressively Therefore, ideal errand designation changes Gain must surpass expense of relocation Resources utilized by long-running projects may be recovered by proprietor

Slide 18

Node A Node B PVM undertaking 1 PVMD A PVMD B Node C PVM assignment 2 PVMD C Dynamite Initial State Two PVM undertakings conveying through a system of daemons Migrate assignment 2 to hub B

Slide 19

Node A Node B New setting PVM assignment 1 PVMD A PVMD B Node C Program PVM Ckpt PVMD C Prepare for Migration Create new setting for undertaking 2 Tell PVM daemon B to expect messages for assignment 2 Update steering tables in daemons (first B, then A, later C)

Slide 20

Checkpointing Node A Node B New setting PVM assignment 1 PVMD A PVMD B Node C Program PVM Ckpt PVMD C Send checkpoint sign to assignment 2 Flush associations Checkpoint undertaking to plate

Slide 21

Cross-bunch checkpointing (outline) Node A Node B Helper assignment PVM errand 1 PVMD A PVMD B Node C Program PVM Ckpt PVMD C Send checkpoint sign to undertaking 2 Flush associations, close records Checkpoint assignment to circle by means of aide undertaking

Slide 22

Restart Execution Node A Node B New PVM assignment 2 PVM errand 1 PVMD A PVMD B Node C PVMD C Restart checkpointed undertaking 2 on hub B Resume interchanges Re-open & re-position documents

Slide 23

Special contemplations Preserve correspondence PVM ought to keep on running as though nothing happened Use area free tending to Open records Preserve open record state

Slide 24

Performance Migration speed to a great extent reliant on the pace of shared record framework and that depends generally on the system NFS more than 100 Mbps Ethernet 0.4 s < T mig < 15 s for 2 MB < size img < 64 MB Communication pace diminished because of included overhead 25% for 1 byte direct messages 2% for 100 KB circuitous messages

Slide 25

Current status: Dynamite Part Checkpointer operational under Solaris 2.5.1 and higher (UltraSparc, 32 bit) Linux/i386 2.0 and 2.2 (libc5 and glibc 2.0) PVM 3.3.x applications bolstered and tried Pam-Crash (ESI) - auto accident reenactments CEM3D (ESI) - electro-magnetics code Grail (UvA) - vast, straightforward FEM code NAS parallel benchmarks BloodFlow MPI and attachment (Univ. of Krakow) libraries accessible Scheduling not yet agreeable

Slide 26

Architecture: Revisited

Slide 27

Design Considerations High Quality presentation High Frame rate Intuitive cooperation Real-time reaction Interactive Algorithms High execution registering and systems administration...

Slide 28

Problem: Time, time what has happened to us?

Slide 29

Solution: Asynchronicity

Slide 30

A cop to control the nonconcurrent forms

Slide 31

Runtime Support Need bland system to bolster modalities Need interoperability High Level Architecture (HLA): information dispersion crosswise over heterogeneous stages adaptable characteristic and proprietorship components propelled time administration

Slide 32

Provoking a bit… Progress in common sciences originates from dismembering things ... Progress in software engineering originates from uniting things...

Slide 33

Proof is in the pudding... Analytic Findings Occluded right iliac conduit 75% stenosis in left iliac supply route Occluded left SFA Diffuse illness in right SFA

Slide 34

Problem: From Image to Simulation MR Scan of Abdomen MR Scan of Legs

Slide 35

Solution: 3DManual introduction Place begin point Place one or more end focuses Wave proliferates from begin to end point Backtrack = first estimation of the centerline Wave spreads from ‘centerline’  vessel divider Distance Transform from vessel divider to focus  centerline

Slide 36

Wavefront Propagation Place begin point Place one or more end focuses Wave engenders from begin to end point Backtrack = first estimation of the centerline Wave proliferates from ‘centerline’  vessel divider Distance Transform from vessel divider to focus  centerline

Slide 37

MRA: Backtrack Place begin point Place one or more end focuses Wave engenders from begin to end point Backtrack = first estimation of the centerline Wave spreads from ‘centerline’  vessel divider Distance Transform from vessel divider to focus  centerline

Slide 38

MRA: Wavefront Propagation Place begin point Place one or more end focuses Wave proliferates from begin to end point Backtrack = first estimation of the centerline Wave engenders from ‘centerline’  vessel divider Distance Transform from vessel divider to focus  centerline

Slide 39

MRA: Distance Transform Place begin point Place one or more end focuses Wave spreads from begin to end point Backtrack = first estimation of the centerline Wave engenders from ‘centerline’  vessel divider Distance Transform from vessel divider to focus  centerline

Slide 40

3-D determination of locale of interest

Slide 41

Tracking the vessels

Slide 42

Building the Geometric Models

Slide 43

VR-Interaction

Slide 44

Alternate Treatments Preop AFB w/E-S Prox. Anast. AFB w/E-E Prox. Anast. Angio w/Fem-Fem Angio w/Fem-Fem & Fem-Pop

Slide 45

Problem: Flow through complex geometry After deciding the vascular structure mimic the blood-stream and weight drop… Conventional CFD strategies may fall flat: Complex geometry Numerical insecurity wrt connection Inefficient shear-stress count

Slide 46

Solution to intelligent stream reproduction Use Cellular Automata as a mesoscopic model framework: Simple neighborhood cooperation Support for genuine material science and heuristics Computational proficient

Slide 47

Mesoscopic Fluid Model Fluid model with Cellular Automata rules Collision: particles reshuffle speeds Imposed Constraints Conservation of mass Conservation of energy Isotropy Details...

Slide 48

...Equivalence with NS For cross section with enough symmetry: proportional to the ceaseless incompressible Navier-Stokes mathematical statements: Implicit parallel and complex geometry support.

Slide 49

Efficient Calculation of Shear-Stress Perpendicular force exchange: AND the energy stress tensor P that is directly identified with the shear stresses s stomach muscle From LBE plan:

Slide 50

10 cm/sec 0 cm/sec Velocity Magnitude

Slide 51

Peak Systolic Pressures - Rest 150 mmHg 50 mmHg Preop AFB w/E-S Prox. Anast. AFB w/E-E Prox. Anast. Angio w/Fem-Fem Angio w/Fem-Fem & Fem-Pop

Slide 52

… last slides...

Slide 53

Other Virtual Laboratory Applications @ UvA Computing in Physics Computing in Engineering Computing in Engineering Bio-medicinal Computation Bio-informatics Environment Cultural Inheritance Environment VL for Material Science Traffic Payment for portability Apply VL in non-nature of administration environment Study of blood course through veins DNA Research Art objects safeguarding reclamation Meta information Integration Combining critical thinking & information serious situations Modeling VL in non-QoS circumstance environment Integration of reenactment & perception by man on the up and up Combing information mining & keen information bases Collaborative information incorporation User Central-part Central-part Virtual Laboratory Virtual Laboratory ViSE ComCol VIMCO Physical Apparatus Internet and Web Software Internet and Web Software Distributed Computer base Distributed Computer base

Slide 54

Acknowledgments RUL/AZL: H.

Recommended
View more...