28
pages
English
Documents
Le téléchargement nécessite un accès à la bibliothèque YouScribe Tout savoir sur nos offres
28
pages
English
Ebook
Le téléchargement nécessite un accès à la bibliothèque YouScribe Tout savoir sur nos offres
CS161: Operating Systems
Matt Welsh
mdw@eecs.harvard.edu
Lecture 24: Distributed Operating Systems and Amoeba
May 3, 2007
© 2007 Matt Welsh – Harvard University 1Distributed Operating Systems
So far we have been talking about an OS that runs on a single
machine
What about an OS that runs across many machines,
distributed on a network?
Why might such a thing be useful?
●
●
●
© 2007 Matt Welsh – Harvard University 2Distributed Operating Systems
So far we have been talking about an OS that runs on a single
machine
What about an OS that runs across many machines,
distributed on a network?
Why might such a thing be useful?
● Transparent access to remote resources
● Support parallel computing – running programs across many machines simultaneously
● Provide a single system image across many different machines
Today: The Amoeba Distributed Operating System
● One of the first and most influential distributed OS projects
© 2007 Matt Welsh – Harvard University 3The rise of distributed computing
Mid-70's: Ethernet developed
● Developed by Bob Metcalfe while at Xerox PARC in 1973, went on to found 3Com
Allows machines to be connected into a local area network
User on one machine can access services and data on other machines
© 2007 Matt Welsh – Harvard University 4Internetworking
Routing allows multiple local networks to communicate:
140.247.62.xxx
140.247.60.xxx
140.247.62.1
140.247.60.1
Router (or “gateway”)
140.247.2.1
140.247.2.xxx
© 2007 Matt Welsh – Harvard University 5Internetworking
OK, but how do we communicate over long distances?
Harvard MIT
Autonomous system
(AS)
Cogent
Backbone router
Sprint
BUVerizon
● Long-distance (WAN) links between multiple sites
● So now, a user at Harvard can use services or data anywhere on the network!
© 2007 Matt Welsh – Harvard University 6Example Applications
Remote access to supercomputing resources
● User at University A can run jobs on supercomputer at University B
● e.g., Simulations of galaxy formation, biomedical data processing, etc.
Remote access to large data sets
● e.g., Archived scientific data, satellite images, genomic data, etc.
© 2007 Matt Welsh – Harvard University 7Grid Computing
Modern concept: Tie all of the world's high-end computers together
into a massive computing grid
© 2007 Matt Welsh – Harvard University 8http://tg-monitor.ncsa.teragrid.org/
© 2007 Matt Welsh – Harvard University 9Cluster Computing
Network of (generally identical) computers
Provide many fast CPUs for parallel
computing applications
Generally uses very fast local area network
● Asynchronous Transfer Mode (ATM)
● FiberChannel
● InfiniBand
● Myrinet
● These days: 10 Gbit Ethernet
Amoeba cluster at Vrije Universitat
© 2007 Matt Welsh – Harvard University 10