a p2p distributed computing platform: korea@home 2004.1.28(wed) c.y. ‘connor’ park...
Post on 18-Dec-2015
217 views
TRANSCRIPT
A P2P Distributed A P2P Distributed Computing Platform:Computing Platform:
Korea@HomeKorea@Home
2004.1.28(Wed)2004.1.28(Wed)C.Y. ‘Connor’ ParkC.Y. ‘Connor’ Park
[email protected]@kisti.re.krKISTI Supercomputing CenterKISTI Supercomputing Center
22
OverviewOverview
Korea@HomeKorea@Home ( (KatHKatH) : A Software ) : A Software Platform for Distributed Computing Platform for Distributed Computing using Volunteer Computer Resourcesusing Volunteer Computer Resources
FeaturesFeatures Hybrid Peer-to-Peer architectureHybrid Peer-to-Peer architecture Secure PKI-based Security and Digital Secure PKI-based Security and Digital
SignatureSignature Web Services with SOAP messagingWeb Services with SOAP messaging
33
MotivationMotivation
To take advantage of and stimulate To take advantage of and stimulate ‘Broadband Network Infrastructure in ‘Broadband Network Infrastructure in Korea’Korea’
To meet the growing needs of To meet the growing needs of computing power for applicationscomputing power for applications
Funded by ‘Ministry of Information Funded by ‘Ministry of Information and Communication (MIC)’and Communication (MIC)’
44
Broadband Network Broadband Network Infrastructure in KoreaInfrastructure in Korea
Over 10M Broadband network (xDSL, Cable) Over 10M Broadband network (xDSL, Cable) households (over 70% penetration) households (over 70% penetration) – 2002.11– 2002.11
Over 58.4% of PCs with Pentium3+Over 58.4% of PCs with Pentium3+
76.9 78.6
23.1 21.4
0%
20%
40%
60%
80%
100%
2001.12 2002.12
Own PCs Not own PCs
82.3 89.3
17.7 10.7
0%
20%
40%
60%
80%
100%
2001.12 2002.12
Internet-enable Internet-unable
55
Similar ProjectsSimilar Projects
ScienceScience SETI@Home, SETI@Home,
climateprediction.netclimateprediction.net
Life ScienceLife Science Folding@Home, Folding@Home,
FightAIDS@HomeFightAIDS@Home
CryptographyCryptography Distributed.net, Distributed.net,
RSAttack576RSAttack576
MathematicsMathematics Mersenne Prime Search, Mersenne Prime Search,
Pi(x) ProjectPi(x) Project
InternetInternet Grub, CycleTraderGrub, CycleTrader
FinancialFinancial MoneyBeeMoneyBee
GamesGames ChessBrainChessBrain
MiscellaneousMiscellaneous DALiWorldDALiWorld
from http://www.aspenleaf.com/distributed/index.html
66
Project HistoryProject History 2002. 4. 2002. 4.
MIC MIC launchedlaunched the pilot project ‘Korea@Home’the pilot project ‘Korea@Home’ Application : Virtual Screening for New Drug DesignApplication : Virtual Screening for New Drug Design
2002. 8.2002. 8. First pilot productionFirst pilot production and test and test Client/Server Architecture ModelClient/Server Architecture Model
2002. 12.2002. 12. Second pilot productionSecond pilot production and test and test
2003. 4.2003. 4. Regularize the project “P2P Internet-based Distributed Regularize the project “P2P Internet-based Distributed
Computing Platform Development”Computing Platform Development” 2003. 9.2003. 9.
Announcement of Hybrid P2P-style Distributed Announcement of Hybrid P2P-style Distributed Computing PlatformComputing Platform
77
ArchitectureArchitecture
CooperationCooperation
Cooperation
InternetInternet
Large-scaleApplication
ResourceManagement
Subscription
Distribution
Distribution Distribution
Distribution
Agent
Agent Agent
Agent
Job Request
88
Job DistributionJob Distribution
InternetInternet
KatH Server
Agent Agent Agent
Agent
Authentication
Working Specification Request• Is there any job for
me to work?•Who has the job
data? Getting data,
working, and returning result to server
99
2002 Pilot Projects2002 Pilot Projects
Application : Virtual Screening for New Application : Virtual Screening for New Drug DesignDrug Design
Client-Server ArchitectureClient-Server Architecture An AgentAn Agent
Receives about 400KB-size fileReceives about 400KB-size file Executes during 20 minutes per working unitExecutes during 20 minutes per working unit Returns about 190KB-size fileReturns about 190KB-size file
Total number of working job units : 50,000Total number of working job units : 50,000
1010
Virtual ScreeningVirtual Screening
Structure Analysis
Target Protein
Share
Analysis
Chemical DB
Disease Protein Ligand DB
Server
+ +
Client 1
+ +
+ +
Client 2
Client 3
Virtual High-Throughput Screening
Internet
Calculating Ligand PMI
Looking forDrug Candidate
Clinical Demo New Drug Mass-produce
Calculating Active Site PMI
1111
Experimental ResultsExperimental Results 1st Test1st Test
1,217 installed agents1,217 installed agents 2002.8.22-9.2 2002.8.22-9.2 Max Max 53.553.5 Gflops / Avg. Gflops / Avg. 31.7831.78 Gflops Gflops
2nd Test2nd Test 6,472 installed agents6,472 installed agents 2003.2.6-2.142003.2.6-2.14 Max Max 644644 Gflops / Avg. Gflops / Avg. 470.6470.6 Gflops Gflops
※ ※ World Top Rank 124 - 629.7 Gflops: World Top Rank 124 - 629.7 Gflops: IBM xSeries Cluster Xeon 2.8 GHz - Myrinet/ 192IBM xSeries Cluster Xeon 2.8 GHz - Myrinet/ 192((from from http://www.top500.orghttp://www.top500.org, 2003.6, 2003.6))
1212
Intention of ParticipationIntention of Participation
Korea@Home Understanding Intention of Providing Resources
Well-Understanding (44.9%)
Understanding roughly (21.8%)
Don’t Know (33.3%)
Keep up Providing (59.5%)
Providing later (30.4%)
Providing with Compensation (7.2%)
Don’t (2.9%)
•67.7% of Netizen already know what is Distributed Computing and Korea@Home.
•89.9% of Netizen have intention of providing their PC resources.
1313
2003 Progresses2003 Progresses
P2P data distributionP2P data distribution New ApplicationsNew Applications
Virtual Screening for New Drug DesignVirtual Screening for New Drug Design Global Risk Management for financial Global Risk Management for financial
investmentinvestment 3D image rendering3D image rendering
Supplementary plug-inSupplementary plug-in Remote backup serviceRemote backup service
1414
FutureworkFuturework
Make P2P agents more intelligent and Make P2P agents more intelligent and cooperative with each other.cooperative with each other.
A set of APIs for application developersA set of APIs for application developers Fault-tolerance : checkpointing & Fault-tolerance : checkpointing &
rollback recoveryrollback recovery IPv6 versionIPv6 version Open-Source (?)Open-Source (?) Windows-only Windows-only Unix version Unix version
1515
Thanks a lot!Thanks a lot! ^^ ^^
[email protected]@kisti.re.kr, , http://www.koreaathome.orghttp://www.koreaathome.org
1616
Design GoalsDesign Goals
Application Developers (Job Application Developers (Job Requester)Requester) ConvenienceConvenience
They should be able to use easily.They should be able to use easily.
FlexibilityFlexibility The platform is general enough to be used The platform is general enough to be used
by various applications.by various applications.
CorrectnessCorrectness The platform should return correct results.The platform should return correct results.
1717
Design GoalsDesign Goals
PC agents (Resource Providers)PC agents (Resource Providers) TransparencyTransparency
PC users are not disturbed in their own PC users are not disturbed in their own works.works.
ReliabilityReliability It’s reliable not to make any security and It’s reliable not to make any security and
privacy problem.privacy problem.
RewardReward There is any kind of compensation, even just There is any kind of compensation, even just
in mind.in mind.
1818
Design GoalsDesign Goals
PlatformPlatform ScalabilityScalability
Performance should be improved according Performance should be improved according to the growth of agents.to the growth of agents.
AdaptabilityAdaptability It should put up with the dynamic features of It should put up with the dynamic features of
Internet.Internet.
1919
Implementation ChoicesImplementation Choices
Web ServicesWeb Services SOAP messaging over httpsSOAP messaging over https
Request-Response ModelRequest-Response Model It is Agents’ duty to make most decisions.It is Agents’ duty to make most decisions.
PKI-based certification and Digital PKI-based certification and Digital Signature for filesSignature for files
Make server as simple as possible, and Make PC agents
more intelligent
2020
Functional ClassificationFunctional Classification Resource ManagementResource Management
Authentication, CertificationAuthentication, Certification Simple accountingSimple accounting
Scheduling and Job distributionScheduling and Job distribution Job distribution with hybrid P2P styleJob distribution with hybrid P2P style Result gatheringResult gathering
Client ManagementClient Management Client authenticationClient authentication Application preparation to workApplication preparation to work Digital signingDigital signing Result returning to clientResult returning to client
2121
Networking IssuesNetworking Issues
Network SecurityNetwork Security DoS, Spoofing, Sniffing, …DoS, Spoofing, Sniffing, …
Variable ConnectionVariable Connection Dial-up, cable, xDSL, T1/E1, …Dial-up, cable, xDSL, T1/E1, … IPv6, WiFi, Bluetooth, …IPv6, WiFi, Bluetooth, …
Network UnobtrusivenessNetwork Unobtrusiveness Background/shared communicationBackground/shared communication
Transparency (Firewall/NAT)Transparency (Firewall/NAT)