Dad i want a supercomputer on my next
Download as PPT, PDF1 like466 views
This document discusses Beowulf clusters, which are low-cost high-performance computing systems built from commodity off-the-shelf computers. It provides details on a specific Beowulf cluster built at Caltech in 1996 using 16 Pentium Pro processors that achieved a total of 1.25 billion floating-point operations per second (Gflops) at a much lower cost than conventional supercomputers. It also outlines the benefits of computational modeling and simulation using Beowulf clusters and provides steps for building and programming a basic Beowulf cluster.
Dad i want a supercomputer on my next
- 1. Beowulf Cluster, building cheap powerful hpc By Akash Sahoo
- 3. Speed •16 Pentium Ppro 200Mhz each •2GB Memory •1.25 Glops •130 Mflops individual speed •Made by Caltech in ‘96 FLOPS (or flops or flop/s, for floating-point operations per second) is a measure of a computer's performance, especially in fields of scientific calculations. It is instruction per second i7-920 has 40Gflops In 2007 a Microwulf Cluster (less nodes) with just 4 nodes gave a performance of 26.25 Gflops (Wikipedia)
- 4. Why Speed Computer simulations are far cheaper and faster than physical experiments Computers can solve a much wider range of problems than specific laboratory equipment can Computational approaches are only limited by computer speed and memory capacity, while physical experiments have many practical constraints Uses: Predictive Modeling and Simulations, Engineering Design and Automation, Military, Visualization
- 5. Why Beowulf Commodity Hardware, No need to throw old PCs !! Ease of addition of nodes Work if nodes fail VVVV Cheap (Supercomputer costs million) World Fastest is a cheap Beowulf Cluster Can be used in Homes, educational Institutes Students will study and write parallel, optimized code Simulation Weather, BLAST (comparing primary biological sequence information), BIONIC (Cancer), distributed Matlab, DNA, Image processing, Web server, own software Software can run on one/many node as specified INDIA has outdated supercomputer - PARAM
- 6. Parallel Computing Shared Memory NFS Message Passing Efficiency Latency Topology Admin and maintainace Redundancy Workload, Unbalancing, Arch Energy, cost, scalable
- 7. Building It Install Linux in each node (MPICH2), Can be build on Windows (PVM) SSH Equivalence SSH Hostkeys – RSA, DSA Installation of PDSH Setting NFS Creation of MPI users Installation of compat-g77,gcc MPI install, LINPACK Benchmark OSCAR, ROCKS, SCARY- maintainance
- 8. Mpdboot –n 3 Starting cluster Mpdtrace Know which nodes are running
- 9. Mpicc Compiling prog Mpiexec –n 1 ./prog Execute a sample prog on 1 node
- 10. Same prog on 2 nodes Same prog on 3 nodes
- 11. Improvements Add fibre optic cable Dual Lan card Powerful Gateway, Router Faster computers – Cost High, Performance High Keep in mind-Total memory, Memory bandwidth, Storage bandwidth, System interconnect latency, The effect of interconnect topologies such as meshes, cubes, and fat trees on application design SSD Harddisk, RAID, Graphics Card
- 12. Notes Gave a performance boost of > 200% on 3 nodes Originality : Security both RSA, DSA. PDSH makes run commands in all nodes at once Commercialization: giving manuals on prog, software, setting cluster, maitainace Proj done in 3rd year
- 13. Patents Dynamically allocated cluster system Charles Evert Prael ; (Patent No: US5093912) The invention provides a system and method for managing clusters of parallel processors for use by groups and individuals requiring supercomputer level computational power. Given Scheduling algo PSB, job queue… 28 claims mostly on scheduling Method and apparatus for securing and managing cluster David Bruce Kumhyr 2002