Some notes on Jon “Maddog” Hall’s session Extreme Linux Programming – A Continuum.
It was new to me when Jon told us that the Titanic movie used 160 alpha processors with Linux to render the movie. The final rendering took about a year. The producers saved 500 000$ compared to proprietary solutions, a circumstance that Jon commented with: “So the world’s most xpensive movie was half a million dollar cheaper”.
Jon examplified how Linux is used for super computing when finding quarks (physics), doing adaptive control of earthquakes, simulating meteorits crashing New York, mammograms (breast cancer) analysis.
In these cases, Jon said, Linux helps with its cost efficiency, because often people say: “We know how to solve the problem, but we cannot afford to solve it.”, until they see the cost benefits of using Linux for super computing.
Jon drew the following future and past chronologic line, showing “Where does Linux belong?”:
– Beowulfs 1994/1995
– Small-Mid Range Servers 1998
– Embedded Systems 2000
– Commodity based NUMA machines 2003
– Desktop 2003/2004
And the nice thing is, he added, that all of it is based on one set of APIs.
Linux is just perfect for super computing, he said, because “the ntworking is built in” and “parallelism screems at you”. With Linux, you have parallelism even in single-cpu machines where it cuts down on I/O wait time and keeps memory and cache “warmer”.
The investment protection that Linux offers to super computing implementation, are based on the:
– standard operating system
– standard architectures
– standard programming techniques
inherent to Linux.
Oh, and I learned a new acronym: RAS = Reliability/Availability/Scalability.