Grid Computing Mr. Tim Stitt Dept. Computer Science and Mathematics University of the West Indies Talk Outline 1. The Data Challenge 2. Solutions to the Data Challenge 3. What is the Grid ? 4. How will the Grid Work ? 5. Grid Applications The Data Challenge After fifty years of innovation: The raw speed of individual computers has increased by a factor of around one million: yet they are still too slow for many challenging scientific problems. Grand Challenge Problem Detectors at the Large Hadron Collider at CERN, Geneva will be producing several Petabytes of data per year - a million times the storage capacity of the average desktop computer; - accounts for nearly 10% of all the information produced by humans each year. Performing the most rudimentary analysis of the LHC data will require close to 20 TeraFlops ( a trillion floating-point operations per second ) The fastest contemporary supercomputer just manages close to 13 Teraflops of computing power. ASC Q • 4,096 Alpha EV-68 processors ( 1.25 GhZ each ) • 33 TB of memory • 664 TB of disk space • 13 Teraflops (peak) • >football field • >1 Mwatt • >$100M Supplier: Client: Main Application: IBM USA Department of Energy Simulated Testing of Nuclear Weapons Stockpile Solutions to the Data Challenge ? Nearly every organization is sitting on top of enormous unused computing capacity: Mainframes idle 40% of the time; Unix servers are actually “serving” less than 10% of the time; Most PC’s do nothing for 95% of a typical day. Consider a hotel with 95% vacancy or an airline with 90% of its fleet on the ground. Solutions to the Data Challenge ? Can we use this to our advantage ? Answer: Yes Large Hadron Collider, CERN, Geneva Solutions to the Data Challenge ? Several approaches to harnessing idle compute time and resources have been implemented over the last decade: 1. Distributed Computing 2. Metacomputing 3. Linking computers over a network Linking Supercomputers with high-speed networks Cluster Computing Overcome the need for expensive commercial supercomputers by linking together commodity PC’s – good scalability. Solutions to the Data Challenge ? 4. Peer-to-Peer Computing 5. Individuals download software to their local machines and make it public to other users bypassing a central server e.g. Kazaa Internet Computing ( or cyclescavenging ) 6. Different parts of problems can be worked on simultaneously by users who download chunks of data over the Internet, process the data and return their results. Based on goodwill and hence not a viable strategy for all tasks e.g. SETI project. The Grid NorduGrid Scalable Computing P E R F O R M A N C E 2100 2100 2100 2100 2100 2100 2100 2100 2100 + Q o S Personal Device SMPs or SuperComputers Local Cluster Enterprise Cluster/Grid Global Grid Inter Planet Grid What is the Grid ? Internet Computing, clustering, meta-computing etc. are all special cases of something much more powerful which includes: the ability for communities to share resources to tackle common goals. Science and business today are increasingly: collaborative and multidisciplinary in nature; span institutions, states, countries and continents. What is the Grid ? Email and the Web are basic mechanisms to allow such groups to work together but what could happen if we could link: 1. 2. 3. 4. 5. Data and Databases Computers Sensors Handhelds Radio Telescopes … into a single virtual laboratory. So what is the Grid ? In a paragraph or two: Whereas the Web is a service for sharing information over the Internet, the Grid is a service for sharing computer power and data storage capacity over the Internet. The Grid goes well beyond simple communication between computers, and aims ultimately to turn the global network of computers into one vast computational resource. Another Definition: A grid is a type of parallel and distributed system that enables the sharing, selection, and aggregation of geographically distributed "autonomous" resources dynamically at runtime depending on their availability, capability, performance, cost, and users' quality-of-service requirements. The Grid The term “Grid” was chosen by analogy with the electric power grid. “The Grid would let users tap processing power off the internet as easily as electric power can be drawn from a wall socket” The Power Grid Electrical Power Grid The “Grid” When you plug in an appliance to a socket, you don’t care where the power came from e.g. wind, coal or a nuclear plant When you sit in front of your computer to solve a problem you know that whatever computer you plug into the Internet, you will get the computing power and storage capacity you need to complete the job The infrastructure is called the “power grid”. It links together power plants of different kinds with your home through transmission stations, power stations, transformers, powerlines etc. The infrastructure is called the “Grid”. It links together computing resources such as PC’s, workstations, servers, storage elements etc. The Power Grid Electrical Power Grid The Grid Is Transperent: Is Transparent: No need to worry about how and where the electricity is generated. You don’t need to know what computer processes you request and where the data is that it needs. Is Pervasive: Is Pervasive: Electricity is available essentially everywhere and can be accessed simply through a standard wall socket Remote computing resources will be accessible from different platforms, including desktops, laptops, PDA’s, mobile phones, through a web browser ( a portal ) Is a Utility: Is a Utility: You ask for electricity, you get it and pay for it You ask for computer power or storage and you get it. You also pay for what you get. This idea of a computing power grid is not new: “The time-sharing computer system can unite a group of investigators …. one can conceive of such a facility as an … intellectual public utility.” - Fernando Corbato and Robert Fano , 1966 “We will perhaps see the spread of ‘computer utilities’, which, like present electric and telephone utilities, will service individual homes and offices across the country.” - Len Kleinrock, 1967 It should be noted that reference to grid computing can presently lead to some confusion: The current reality now and for a while is that there is not one single “Grid” ( and may never be ! ). Instead there are many grids ( or virtual organizations ) evolving: - National Grids > couple high end resources across a nation e.g. the e-Science program national Grid in the U.K. - Private Grids > Local grids for use in institutions such as hospitals, corporations etc. - Project Grids > Grids developed to meet the need of multi-institutional research groups and multi-company virtual teams. NASA’s Information Power Grid - GoodWill Grids > Anyone owning a computer at home can donate some computer capacity to a good cause - Consumer Grids > Resources are shared on a commercial basis rather than goodwill or mutual self-interest. Companies or organisations rent distributed resources from the owners. Gusto How will it work ? Grid development relies on advanced software, called Middleware, which: - Ensures seamless communication between different computers and different parts of the world; - Provides a grid search engine which will not only find the data the user requires but also the data processing techniques and computing power to carry it out; - Distributes the computing task to wherever in the world there is spare capacity and then send the results back to the user. The Grid Middleware ( in more detail ): 1. Finds convienient places for computing tasks to be run 2. Discovers and optimises use of the widely dispersed resources 3. Organises efficient access to scientific data 4. 5. Deals with authentication to the different sites that the user will be using Interfaces to local site authorisation and resource allocation policies 6. Runs the jobs 7. Monitors progress 8. Recovers from problems 9. Tells the user when the work is completed and transfers back the results. The Five Big Ideas Resource Sharing - Challenge: resources owned by many different people Secure Access - 3 A’s: Access, Authentication, Authorization Resource Use - Make use of resources efficiently Death of Distance - High speed networking technology makes Grid possible Open Standards - Coordinated by Global Grid Forum - Agreement on core technologies such as the Globus Toolkit Grid Layered Architecture Natural Language Engineering Molecular Docking High Energy Physics Portfolio Applications Analysis Brain Activity Analysis Application Toolkits GlobusView DUROC MPI Condor-G MPI LSF PBS GSI-FTP GSI MDS Condor Testbed Status Nimrod/G Grid Services Nexus I/O HPC++ Linux globusrun HBM GASS TCP NT Solaris User-Level Middleware (Grid Tools) Core Grid Middleware GRAM Grid Fabric NQE GAMESS Chemistry Grid Apps. UDP DiffServ Grid Fabric Current Grid Applications NCSA Origin Caltech Exemplar CEWES SP Maui SP SF-Express distributed interactive simulation. 100K vehicles (2002 goal) using 13 computers, 1386 nodes, 9 sites. Distributed Molecular Docking Molecules Protein Chemical Databases (legacy, in .MOL2 format) Involves screening millions of chemical compounds (molecules) in the Chemical DataBase (CDB) to identify those having potential to serve as drug candidates. Medical Healthcare Applications • Digital image archives • Collaborative virtual environments • On-line clinical conferences “The Grid will enable a standardized, distributed digital mammography resource for improving diagnostic confidence" “The Grid makes it possible to use large collections of images in new, dynamic ways, including medical diagnosis.” “The ability to visualise 3D medical images is key to the diagnosis of pathologies and presurgical planning” Nanotechnology • New and 'better' materials • Benefits in pharmaceuticals, agrochemicals, food production, electronics manufacture from the faster, cheaper discovery of new catalysts, metals, polymers, organic and inorganic materials “The Grid has the potential to store and analyze data on a scale that will support faster, cheaper synthesis of a whole range of new materials.” Natural Resources/Environments • Modeling and prediction of earthquakes • Climate change studies and weather forecast • Pollution control • Socio-economic growth planning, financial modeling and performance optimization “Federations of heterogeneous databases can be exploited through the Grid to solve complex questions about global issues such as biodiversity.” myGrid Project - Bioinformatics Imminent ‘deluge’ of genomics data - Highly heterogeneous, Highly complex and interrelated Convergence of data and literature archives - Database access from the Grid - Process enactment on the Grid - Personalisation services - Metadata services National Virtual Observatory http://virtualsky.org/ from Caltech CACR Caltech Astronomy Microsoft Research Virtual Sky has 140,000,000 tiles 140 Gbyte Change scale Change theme Optical (DPOSS) Xray (ROSAT) theme Coma cluster Belle Particle Physics Experiment A Running experiment based in KEK B-Factory, Japan Investigating fundamental violation of symmetry in nature (Charge Parity) which may help explain the universal matter – antimatter imbalance. Collaboration 400 people, 50 institutes 100’s TB data currently UoM School of Physics is an active participant and have led the Grid-enabling of the Belle data analysis framework. Famous Predictions on Technology “The world will only need five computers” Thomas.J.Watson IBM “There is absolutely no need for a computer in the home.” Ken Olsen, DEC “640 kilobytes is all the memory you will ever need.” Bill Gates, Microsoft Announcement "High-Performance Computing, Parallelism and Applications" Summer School July 5-10, 2004 Centre Commun de Calcul Intensif (C3I) Fouillole Campus de l'Université des Antilles et de la Guyane.