Handling complexity and change in grid computing

1
Handling complexity and change in grid computing

• Gosia Wrzesinska

2
Computers are useful
3
Computers are useful

• Computers can solve many useful problems
• But some problems are so complex that solving
  them would take many years
• Weather prediction
• Decoding human genome
• Breaking cryptographical keys
• Speech recognition

4
What can we do?

• Buy a supercomputer
• (expensive)
• Build a supercomputer out of a lot of (cheap)
  computers

5
The world is a supercomputer

• There are a lot of computers in the world
• Many of them connected by the Internet
• Why not use them as a huge supercomputer?
• We call it a grid

6
Programming grid applications is hard

• Computers might be far away from each other
• Communication is slow
• Application must tolerate wide-area latencies

15 000 km
7
Programming grid applications is hard

• Computers can crash
• Computers can be switched off by their owners
• New computers can be added
• Application needs to be fault-tolerant (tolerate
  crashes) and malleable (able to expand to more
  computers or shrink)

8
Programming grid applications is hard

• Grids are dynamic
• Computers come and go
• Computers are sometimes idle but sometimes very
  busy
• Network conditions change
• Application needs to
• adapt like a living
• organism

9
So, what is the thesis about?

• The goal of the thesis is to simplify writing
  grid applications
• Provide a set of tools (compilers, software
  libraries) that make programming easier by
  solving some of the problems
• We call it a grid programming environment

10
Contributions

• Using our grid programming environment we
  implemented a number of applications
• Astrophysical simulation
• Genome sequence alignment
• VLSI design
• And we have shown that they
• Achieve good performance on the grid
• Can tolerate crashing nodes
• Can adapt to changing conditions

11
Thank you.