Slide 1

Slide 1 text

Making free science free Juan Julián Merelo Guervós Free Software Office http://osl.ugr.es & GeNeura Team University de Granada @jjmerelo Identi.ca/jjmerelo

Slide 2

Slide 2 text

Free science free | @jjmerelo 2 Why?

Slide 3

Slide 3 text

Free science free | @jjmerelo 3 My wishlist ● Low entry threshold ● Massive scalability ● All in one system ● Fault-tolerance (except byzantine failures)

Slide 4

Slide 4 text

Free science free | @jjmerelo 4 (In)Volunteer computing ● BOINC as a paradigm. ● Infraestructure based in PHP/MySQL ● Handling 8 million request/day

Slide 5

Slide 5 text

Free science free | @jjmerelo 5 OK, I buy it. What do I need? ● Tasks ● Users ● Groups ● Statistics ● GUI

Slide 6

Slide 6 text

Free science free | @jjmerelo 6 What do you want it for? ● Low cost or free science ● Citizen science ● Evolutionary algorithm experiments (or any other, for that matter) ● Web science: social distributed computers

Slide 7

Slide 7 text

Free science free | @jjmerelo 7 That's why We need an object store

Slide 8

Slide 8 text

Free science free | @jjmerelo 8 What can we use

Slide 9

Slide 9 text

Free science free | @jjmerelo 9 Software as a service

Slide 10

Slide 10 text

Free science free | @jjmerelo 10 Almost, but not quite ● Usually freemium services (pay after a use level) ● Heroku, OpenShift, Amazon, Google AppSpot... ● There's science in it: optimize the number of requests to the server to avoid charges.

Slide 11

Slide 11 text

Free science free | @jjmerelo 11 Can do nothing without you ● Eventually, somebody has to lend CPU time to an experiment. ● AJAX provides a perfect environment for doing that. ● COMET (server-push) too. ● New tech: WebWorkers, apps. ●

Slide 12

Slide 12 text

Free science free | @jjmerelo 12 Free as in free speech ● The whole process must be transparent: Open Science ● People won't give you cycles if they don't trust you. ● Release early, release often, release all: data, source, papers...

Slide 13

Slide 13 text

Free science free | @jjmerelo 13 What have you done? ● Distributed Computing on Rails: ● Ruby on Rails + AJAX on client = Distributed evolutionary algorithms ● Experiment with up to 1000 clients ● AGAJAJ: same thing (pretty much) with mod_perl.

Slide 14

Slide 14 text

Free science free | @jjmerelo 14 EvoSpace ● UGR + UEX + Tijuana tech. ● Redis + Django based distributed evoluitionary computation engine ● Embedded wetware: valid for interactive evolution ● Now moving to Heroku

Slide 15

Slide 15 text

Free science free | @jjmerelo 15 Doing free science is great But it's better if it's just a choice Looking for partners for EC projects? Contact me [email protected]

Slide 16

Slide 16 text

Making free science free Juan Julián Merelo Guervós Free Software Office http://osl.ugr.es & GeNeura Team University de Granada @jjmerelo Identi.ca/jjmerelo Imagen de sunlight in a jar en http://www.flickr.com/photos/sunlightinajar/1111631 514/in/photostream/

Slide 17

Slide 17 text

Free science free | @jjmerelo 2 Why? Imagen de la izquierda de http://www.flickr.com/photos/cipherswarm/2414578959/ Christopher Bowns, de la derecha de ario_ http://www.flickr.com/photos/ario/4422273371/in/photostrea No todo el mundo puede contar con un supercomputador. La computación voluntaria tiene su origen en el proyecto SETI@Home de análisis de señales procedentes del espacio buscando regularidades.

Slide 18

Slide 18 text

Free science free | @jjmerelo 3 My wishlist ● Low entry threshold ● Massive scalability ● All in one system ● Fault-tolerance (except byzantine failures) Foto subida por Texas_mustang a Flickr en http://www.flickr.com/photos/astros/3977731447/in/photostr

Slide 19

Slide 19 text

Free science free | @jjmerelo 4 (In)Volunteer computing ● BOINC as a paradigm. ● Infraestructure based in PHP/MySQL ● Handling 8 million request/day Dato del trabajo de D. B. Anderson, http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=157 Imagen de http://www.flickr.com/photos/23497193@N07/2870438187/ por frseti

Slide 20

Slide 20 text

Free science free | @jjmerelo 5 OK, I buy it. What do I need? ● Tasks ● Users ● Groups ● Statistics ● GUI Imagen de UnitedBOINC http://www.unitedboinc.com/en/website-info/39-site-info/97- Imagen del esquema de BOINC http://bof.linuxforum.dk/2005/slides/Carlos/tappingthematrix

Slide 21

Slide 21 text

Free science free | @jjmerelo 6 What do you want it for? ● Low cost or free science ● Citizen science ● Evolutionary algorithm experiments (or any other, for that matter) ● Web science: social distributed computers Imagen de StudioTempura en http://www.flickr.com/photos/zero101/4783232763/in/photo

Slide 22

Slide 22 text

Free science free | @jjmerelo 7 That's why We need an object store Imagen de Michael Flick en http://www.flickr.com/photos/17773534@N03/3237000861/ Object Store en general se define por oposición a una base de datos tradicional, en la que las columnas son fijas y se estructuran en tablas que tienen relación unas con otras. Una object store, o key-value store, o base de datos NoSQL, almacena documentos indexados, habitualmente, por una clave.

Slide 23

Slide 23 text

Free science free | @jjmerelo 8 What can we use

Slide 24

Slide 24 text

Free science free | @jjmerelo 9 Software as a service There are many free services out there. Picture by Steve Rhodes http://www.flickr.com/photos/ari/5281508652

Slide 25

Slide 25 text

Free science free | @jjmerelo 10 Almost, but not quite ● Usually freemium services (pay after a use level) ● Heroku, OpenShift, Amazon, Google AppSpot... ● There's science in it: optimize the number of requests to the server to avoid charges.

Slide 26

Slide 26 text

Free science free | @jjmerelo 11 Can do nothing without you ● Eventually, somebody has to lend CPU time to an experiment. ● AJAX provides a perfect environment for doing that. ● COMET (server-push) too. ● New tech: WebWorkers, apps. ●

Slide 27

Slide 27 text

Free science free | @jjmerelo 12 Free as in free speech ● The whole process must be transparent: Open Science ● People won't give you cycles if they don't trust you. ● Release early, release often, release all: data, source, papers...

Slide 28

Slide 28 text

Free science free | @jjmerelo 13 What have you done? ● Distributed Computing on Rails: ● Ruby on Rails + AJAX on client = Distributed evolutionary algorithms ● Experiment with up to 1000 clients ● AGAJAJ: same thing (pretty much) with mod_perl.

Slide 29

Slide 29 text

Free science free | @jjmerelo 14 EvoSpace ● UGR + UEX + Tijuana tech. ● Redis + Django based distributed evoluitionary computation engine ● Embedded wetware: valid for interactive evolution ● Now moving to Heroku

Slide 30

Slide 30 text

Free science free | @jjmerelo 15 Doing free science is great But it's better if it's just a choice Looking for partners for EC projects? Contact me [email protected]