Distributed Computing Utilize unused PC resources q Processing
Distributed Computing Utilize unused PC resources q Processing n n Complex calculations Load distribution 25% of storage is unused q q SANs 100 computers 80 gb drives = 6 tb unused
Process Sharing Applications n n n For large-scale computations Data analysis, data mining, scientific computing Research Problems q q q q SETI@Home Folding@Home distributed. net Genome@Home Fight. AIDS@Home climate simulation Economics medicine
Distributed Computing n n n P 2 P is not distributed computing; similar challenges and issues from: sharing and taking advantage of resources available at endpoints and harnessing their power for computationally intensive problems SETI@home, fightaids@home, genome@home Grid computing and e-science q q Computational grids to solve/simulate real-life problems E-Science Commercial applications United Devices, Entropia, Avaki, etc.
Distributed Computing A central coordinator schedules tasks n on volunteer computers, Master worker paradigm, Cycle stealing n Client application Params. /results. Dedicated Applications q q Production applications q Coordinator q Parameters q n Internet Volunteer PC q Volunteer PC Javelin, Bayanihan, JET, Charlotte (based on Java), Commercial Platforms q Volunteer PC Downloads and executes the application Folding@home, Genome@home, Xpulsar@home, Folderol, Exodus, Peer review, Research Platforms q n SETI@Home, distributed. net, Décrypthon (France) q Entropia, Parabon, United Devices, Platform (AC)
Cycle Sharing Model n n Chunks of data are sent to client in suspend mode Data is processed by clients when client is not in use and returned to the master Internet-based (Master-slave) computing Example: SETI@Home scans radio telescope images Master Raw Data Processed Data Bob Ted Carol Alice Data Crunching
SETI@HOME • Launched In 1996 • Scientific experiment - uses Internet-connected computers in the Search for Extraterrestrial Intelligence (SETI) • Distributes a screen saver–based application to users • Applies signal analysis algorithms different data sets to process radio-telescope data. • Has more than 3 million users - used over a million years of CPU time to date SETI@Home Main Server 3. SETI client gets data from server and runs Client/Server P 2 P 1. Install Screen Server 4. Client sends results back to server Radio-telescope Data 2. SETI client (screen Saver) starts
Distributed Computing: SETI@home Search for Extraterrestrial Intelligence that has over two million computers crunching away and downloading data gathered from the Arecibo radio telescope in Puerto Rico The SETI@Home project is widely regarded as the fastest computer in the world In fact, the project has already performed the single largest cumulative computation to date From the architecture point of view Seti@Home is based upon client-server The centralised servers hold enormous amounts of data gathered from the Arecibo radio telescope "listening" to the skies That data needs to be analysed for distinct or unusual radio waves that might suggest extraterrestrial communications http: //setiathome. ssl. berkeley. edu
SETI@Home q Search for Extraterrestrial Intelligence
Processing n Intel’s Netbatch q 10, 000 workstations over 25 locations q Chip design q Shortened time for chip development q Reduced outlay for new mainframes q $500 million savings
Processing n Amerada Hess q q Connects 200 Dell PCs to handle complex seismic data interpretation Allowed them to replace a pair of IBM supercomputers. “We’re running seven times the throughput at a fraction of the cost. ” Richard Ross, CIO
Storage n Intel q q Distribution of computer-based training Prevents large downloads from central servers Preserves bandwidth Preserves expensive network storage
P 2 P Distributed Computing Allows any node to play different roles (client, server, system infrastructure) Client (PC) request PC Client (PC) result PC PC result request Server (PC) accept PC PC PC P 2 P system PC provide Potential communications for parallel applications accept PC Server (PC) PC provide Request may be related to Computations or data Accept concerns computation or data A very simple problem statement but leading to a lot of research issues: scheduling, security, message passing, data storage Large Scale enlarges the problematic: volatility, confidence, etc.
“Three Obstacles to Making P 2 P Distributed Computing Routine” 1) New approaches to problem solving q Data Grids, distributed computing, peer-to-peer, collaboration grids, … 2) Structuring and writing programs q Abstractions, tools Programming Problem 3) Enabling resource sharing across distinct institutions q Resource discovery, access, reservation, allocation; authentication, authorization, policy; communication; fault detection and notification; …Systems Problem Credit: Ian Foster
P 2 P for Distributed Computing or Web Computing n n n The distributed computing P 2 P applications are highlighted by the use of millions of Internet clients to analyze data looking for extraterrestrial life (SETI@home http: //setiathome. ssl. berkeley. edu/ ) and the Newer project examining the folding of proteins ( Folding@home http: //www. stanford. edu/group/pandegroup/Cosm/ ). These are building distributed computing solutions for a special class of applications: q Those that can be divided into a huge number of essentially independent computations, and a central server system doles out separate work chunks to each participating client. q In the parallel computing community, these problems are called "pleasingly or embarrassingly parallel". This approach is included in the P 2 P category because the computing is Peer based even though it does not have the "Peer only communication" characteristic of all aspects of Gnutella and Napster for information transfer. SETI@home and Folding@home are elegantly implemented as screen savers that you download.
P 2 P space: Distributed Computing n Distributed Collaboration q q n Use under utilized Internet and/or network resources for improving computation and data analysis Meta. Computing, Care. Science, Data. Synapse, Distributed. net, Distributed. Science, Entropia, Parabon, The Open Lab Distributed Search Engines q q Used to easily lookup and share files and offer content management Bear. Share, Filetopia, Hotline Connect, Infra. Search, Plebio, Jibe, Lime. Wire, Music. Brainz. org, Neuro. Grid, Next. Page, Redfoot, Opencola, Project Pandango
Entropia Financial Modeling I
Entropia Financial Modeling II n n n Each basic financial instrument can be calculated independently Central Server interprets the total simulation Make Money or Learn what causes market swings or ….
Drug Structure Simulations
United Devices also does Drug Simulation n Parameter Study: do billions of simulations – each with different parameters Search Engine like interface to simulation Works as each calculation fits in a PC – a detailed molecular model would usually not do this
Performance of Entropia Network
Server Server Peer to Peer P 2 P “Illusion” among collaborating clients For Napster like Services or Collaboration
- Slides: 23