Using dropbox for distributed evolutionary computation

1,328 views

Published on

paper presented at th

Published in: Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
1,328
On SlideShare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
6
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide
  • Background picture from http://www.flickr.com/photos/coyote-agile/1578404172/in/photostream/
  • Describe basic working principles of Dropbox as a free file synchronization service, how it monitors certain files and how they are copied when modifed under its own schedule. It's got a permission system that regulates who's got access to which resources; it's usually done per directory, and you can also make directory results publicly available.
  • Every node runs independently, although they are started (roughly) at the same time. After an appointed number of generations, an individual (the best) is logged into the common directory, and another one is taken. A file with the total number of evaluations is also created, and the sum of all evaluations is checked as a termination condition. To speed up processing and reduce overhead speeding up synchronization, the individual is codified into the filename, so that the only thing that is read is the directory, not the content of the file itself. We use different kinds of codifications depending on the chromosome length, but lengths of several hundreds are not a problem. As it can be seen, the fitness is also codified into the name.
  • Picture from Andrew Rivett, Veggiefrog
  • Using dropbox for distributed evolutionary computation

    1. 1. Assessing Speed-ups In Commodity Cloud Storage Services For Distributed Evolutionary Algorithms Maribel García-Arenas, Juan-J. Merelo Antonio M. Mora, Pedro Castillo, Gustavo Romero, JLJ Laredo GeNeura group University of Granada (Spain) Http://geneura.wordpress.com http://twitter.com/geneura
    2. 2. What do clouds smell like?
    3. 3. How can I use cloud storage for distributed evolutionary algoritms? If possible, for free!
    4. 4. We will find out in this paper
    5. 5. What we will do <ul><li>Describe Dropbox, a cloud storage/synchronization system
    6. 6. How we implement a distributed evolutionary algorithm over this system
    7. 7. Raw speedup measures over several computers
    8. 8. Differences between connection modes. </li></ul>
    9. 9. A thunderstorm of experiments <ul><li>1 to 4 different computers.
    10. 10. Gigabit Ethernet or WiFi + WPA connection
    11. 11. Dropbox updated to latest version
    12. 12. MMDP and Trap </li></ul>
    13. 13. Dropbox
    14. 14. Cloudy, with a chance of evolution Single individual Evaluations/node
    15. 15. Trapped in the clouds Ethernet WiFi
    16. 16. MMDP Ethernet WiFi
    17. 17. Conclusions <ul><li>Dropbox can be used for distributing evolutionary algorithms
    18. 18. Load balancing is automatic.
    19. 19. Network matters
    20. 20. Good scaling
    21. 21. Will it mean a better algorithm? </li></ul>
    22. 22. Thanks for your attention Any questions? Http://geneura.wordpress.com Http://twitter.com/geneura

    ×