28
Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Embed Size (px)

Citation preview

Page 1: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Exa-Scale Volunteer Computing

David P. Anderson

Space Sciences LaboratoryU.C. Berkeley

Page 2: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Outline

• Volunteer computing

• BOINC

• Applications

• Research directions

Page 3: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

High-throughputcomputing

High-performancecomputing

program runstoo slow on PC

cluster(MPI)

supercomputer

cluster(batch)

Grid

Commercialcloud

Volunteercomputing

single job

# processors

multiple jobs

10K-1M

1000

100

1

Page 4: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Volunteer computing

• Early projects

– 1997: GIMPS, distributed.net

– 1999: SETI@home, Folding@home

• Today

– ~50 projects

– 500K volunteers

– 1M computers, 2.4M cores

– 10 PetaFLOPS

Page 5: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

The potential of volunteer computing

• The volunteer resource pool• Current PetaFLOPS breakdown:

• Potential: ExaFLOPS by 2010– 4M GPUs * 1 TFLOPS * 0.25 availability

Processor type0

0.51

1.52

2.53

3.54

4.55

4.6

2.4 2.2

1.2

NVIDIA

CPU

PS3 (Cell)

ATI

Page 6: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

BOINC

• Middleware for volunteer computing

– client, server, web

• Based at UC Berkeley Space Sciences Lab

• Open source (LGPL)

• NSF-funded since 2002

• http://boinc.berkeley.edu

Page 7: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

BOINC: volunteers and projects

volunteers projects

CPDN

LHC@home

WCGattachments

Page 8: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

The Utopian vision

• Better research gets more computing power

• An enlightened public decides what’s better

ScientificresearchThe public

resources

education/outreach

Page 9: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Science areas using BOINC• Biology

– protein study, genetic analysis• Medicine

– drug discovery, epidemiology• Physics

– LHC, nanotechnology, quantum computing• Astronomy

– data analysis, cosmology, galactic modeling• Environment

– climate modeling, ecosystem simulation• Math• Graphics rendering

Page 10: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Application types

• Computing-intensive analysis of large data

• Physical simulations

• Genetic algorithms

– GA-based optimization

Page 11: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Climateprediction.net

Page 12: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Einstein@home

• Gravitational waves; gravitational pulsars

Page 13: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

SETI@home

Page 14: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Milkyway@home

Page 15: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

GPUGRID.net

Page 16: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

AQUA@home

• D-Wave Systems

• Simulation of “adiabatic quantum algorithms” for binary quadratic optimization

Page 17: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Quake Catcher Network

Page 18: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Account managers

Page 19: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

BOINC software overview

client

apps

screensaver

GUI

scheduler

MySQL

data server

daemons

volunteer host

project serverHTTP

Page 20: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Client: job scheduling

• Queue multiple jobs

– avoid starvation

– minimize communication

– variety

• Job scheduling

– Round-robin time-slicing

– Earliest deadline first

Page 21: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Client: work fetch policy

• When? From which project? How much?• Goals

– maintain enough work– minimize scheduler requests– honor resource shares

• per-project “debt”

CPU 0

CPU 3

CPU 2

CPU 1

maxmin

Page 22: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

BOINC schedulerapplications

Win32 + NVIDIA

Win64

Mac OS X

app versions

jobs

instances

Win32 N-core

Win32

- HW, SW description- existing workload- per resource type: # of instances requested # of seconds requested

- app version descriptions- job descriptions

Page 23: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Anonymous platform mechanism

• Volunteer supplies app versions.

– security

– optimization

– unsupported platforms

Page 24: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Umbrella projects

Example: IBM World Community Grid

Projectpublicityweb developmentsysadminapp porting

Page 25: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

The Berkeley@home model

• A university has– scientists– a powerful “brand”– PR resources– IT infrastructure– lots of alumni (UCB: 500,000)

Page 26: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Hubs• nanoHUB: “science portal” for nanoscience

– social network + “app store”

– sharing of ideas, data, software

– computational portal

• HUBzero: generalization to other areas

– currently ~20 hubs

• Integration of BOINC with HUBzero

– each hub has a volunteer computing project

Page 27: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Volunteer computing research

• Host characterization• Simulation-based performance study• MPI-type apps• Apps in VMs• Data-intensive computing• Volunteer motivation

Page 28: Exa-Scale Volunteer Computing David P. Anderson Space Sciences Laboratory U.C. Berkeley

Conclusion

• Volunteer computing: Exa-scale potential– GPUs are crucial

• BOINC: enabling technology

• Bottlenecks

– the culture of scientific computing

– organizational models