Session 2
This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA
Overview
Download & View Session 2 as PDF for free.
More details
- Words: 380
- Pages: 5
Concurrent Programming Session 2: Introduction Computer Engineering Department I Iran University U i i off SScience i and dT Technology h l Tehran, Iran
Instructor: Hadi Salimi Distributed Systems Lab. Computer Engineering Department, Department Iran University of Science and Technology, [email protected]
Why Parallelism? y
More complex p computing p g problems ◦ Toyy Storyy animation ◦ Disney and Pixar used 117 SPARC Workstations to render it. ◦ 43 years for a single processor
Why Parallelism? More Complex p Computing p g Problems: y SETI@home y
◦ Analyses data from the world world'ss largest radio telescope, in the hope of detecting signals generated by alien civilizations y
More than 2 million years for a single p computer
Why Parallelism? y
Constructingg Faster Computers p is expensive, hard and impossible: ◦ Silicon technology gy limitations ◦ Cooling problems g ◦ Performance degradation So?
Architectures y
Multi-Processor ◦ A set of processors which communicate through a bus
y
Multi-Computers p ◦ A set of workstations which communicate through g a network.
Spectrum of HW Systems y
MPP ((Massivelyy Parallel Processors)) ◦ Expensive, need HP networking, simultaneous access to shared memory, single address space, hardwired, tightly couples
y
COW (Cluster of Workstations) ◦ Cheaper, easy to make, message passing, not hardwired, flexible, loosely coupled
Architecture y
Discuss about: ◦ ◦ ◦ ◦ ◦ ◦ ◦
The differences of MPs and MCs g g model? Programming Scalability? p g g Ease of programming? Building cost? Hardware availability? Software availability?
Parallelism vs vs. Distribution Sometimes a user’s program which is i h inherently l parallel, ll l can bbe run on a multiprocessor. y This will happen by a parallelizing compiler, proper operating system and specific hardware. y In a distributed system, this will happen among more than a computer. y
Distributed Shared Memory (DSM) Pages of address space distributed among four machines Situation after CPU 1 references page 10 Situation if page 10 is read only and replication is used
Questions y
y y
A complex algorithm is going to be executed. Which one is cheaper? Using a multicomputer or a multiprocessor? What’s the idea behind the DSM? Discuss about COW and MPP from the following point of view: ◦ ◦ ◦ ◦ ◦
Cost, Cost Ease of programming Scalability S d Speed Programming model
Instructor: Hadi Salimi Distributed Systems Lab. Computer Engineering Department, Department Iran University of Science and Technology, [email protected]
Why Parallelism? y
More complex p computing p g problems ◦ Toyy Storyy animation ◦ Disney and Pixar used 117 SPARC Workstations to render it. ◦ 43 years for a single processor
Why Parallelism? More Complex p Computing p g Problems: y SETI@home y
◦ Analyses data from the world world'ss largest radio telescope, in the hope of detecting signals generated by alien civilizations y
More than 2 million years for a single p computer
Why Parallelism? y
Constructingg Faster Computers p is expensive, hard and impossible: ◦ Silicon technology gy limitations ◦ Cooling problems g ◦ Performance degradation So?
Architectures y
Multi-Processor ◦ A set of processors which communicate through a bus
y
Multi-Computers p ◦ A set of workstations which communicate through g a network.
Spectrum of HW Systems y
MPP ((Massivelyy Parallel Processors)) ◦ Expensive, need HP networking, simultaneous access to shared memory, single address space, hardwired, tightly couples
y
COW (Cluster of Workstations) ◦ Cheaper, easy to make, message passing, not hardwired, flexible, loosely coupled
Architecture y
Discuss about: ◦ ◦ ◦ ◦ ◦ ◦ ◦
The differences of MPs and MCs g g model? Programming Scalability? p g g Ease of programming? Building cost? Hardware availability? Software availability?
Parallelism vs vs. Distribution Sometimes a user’s program which is i h inherently l parallel, ll l can bbe run on a multiprocessor. y This will happen by a parallelizing compiler, proper operating system and specific hardware. y In a distributed system, this will happen among more than a computer. y
Distributed Shared Memory (DSM) Pages of address space distributed among four machines Situation after CPU 1 references page 10 Situation if page 10 is read only and replication is used
Questions y
y y
A complex algorithm is going to be executed. Which one is cheaper? Using a multicomputer or a multiprocessor? What’s the idea behind the DSM? Discuss about COW and MPP from the following point of view: ◦ ◦ ◦ ◦ ◦
Cost, Cost Ease of programming Scalability S d Speed Programming model
