L24 - Multiprocessors and Clusters ECS 154B Computer...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
1 ECS 154B Computer Architecture II Winter 2008 Multiprocessors 9.1-9.2 Adapted from slides by Mary Jane Irwin, Penn State 2 Multiprocessors and Clusters • Multiprocessors and clusters connect several processors together to perform computation – Individual processors are inexpensive • Large volume of commercial chips = low cost • Special systems = low volume = high cost – Combined processing power of individual processors • Many applications drive development – Scientific: weather prediction, biological simulation – Web servers, search engines, and databases 3 Basic Organization • Multiprocessors – Several processors connected within a system – Individual processors usually inexpensive – Complex board may increase costs – Maintenance means taking down the entire system • Clusters – Connect several systems through a network – Each node usually inexpensive, but multiplies supporting systems (power supplies, cases, etc.) – Maintenance possible while system is running • Constellation – A cluster of multiprocessors – Compromise of multiprocessors and clusters • Good cost/performance tradeoff • Much of the system can stay running during maintenance 4 Multiprocessor Distribution SMP: A type of multiprocessor, explained shortly MPP: Massively Parallel Processors (Hundreds to Thousands of Processors) SIMD: Single Instruction Multiple Data (Vector Processors) 0 100 200 300 400 500 1 9 93 199 4 5 96 97 98 200 0 2 01 02 03 06 Clusters Constellations SIMDs MPPs SMPs Uniproc's Top 500 Supercomputers (top500.org) #1 IBM BlueGene/L
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
2 5 Main Multiprocessor Questions • Q1: How do parallel processors share data? – Intermediate values and final results – Processing state • Q2: How do parallel processors coordinate? – Algorithm synchronization – Resource sharing • How many processors? – Is the architecture scalable? 6
Background image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 05/06/2008 for the course ECS 154b taught by Professor Krado during the Winter '08 term at UC Davis.

Page1 / 5

L24 - Multiprocessors and Clusters ECS 154B Computer...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online