Cloud infrastructure for GH / Galapagos?

Hi Everybody,

We have developed a strong cloud infrastructure that can run many instances of Grasshopper on demand, bake results, save & access resulting geometry and renders.

We did that for a failed mass customization project we had in mind, and are looking for ideas on how to use this technology in a way that will be beneficial to the GH community.

Our first idea was to implement Galapagos on top of this infrastructure, we can accelerate the process by say, a factor of 20 (this is fluid as we can technically run as many GH instances as we want.). This may open Galapagos to much bigger problems, and of course will solve normal problems much faster. Will that be an interesting project to anyone?

We have invested quite a lot of effort in building this infrastructure, and would be very happy to find a good use for it.. So, any ideas, thoughts and suggestions would be great.

Thanks, Amit.

Replies to This Discussion

Permalink Reply by Andrew Heumann on January 9, 2013 at 3:12pm

I would be very interested in this, although I don't have a specific project in mind. Frankly, I think you should just make the infrastructure available (perhaps on a trial basis) and see what people do with it. Using it to render complex animations for one would be a nice possibility - so that instead of an animation slider where every state is solved in sequence, it could farm out all the frames to be processed in parallel.

Permalink Reply by Ángel Linares on January 9, 2013 at 3:26pm

It looks very interesting :) Perhaps it could be a good start point to share some ideas to let the community to experiment with them...without "touching" and measuring possibilities with the tool in your hands is a little bit complicated to understand how far this cloud system could reach solving problems or optimizing designs.

How many instances are allowed?
Is the problem correctly and automatically divided into clusters or is needed a special user skill/control in that work splitting process?
How are the genome database shared and accessed in every cluster? Could be saved into files?
Is it possible to implement that network over remote Internet connections?
Is it possible to run that cluster using web interfaces?

Permalink Reply by Amit Aviv on January 9, 2013 at 5:24pm

Hi Angel, thanks for the interest.

In terms of hardware infrastructure, the "allowed" number of instances would be in the hundreds, we run on Amazon AWS cloud, and the system we designed is highly scalable.
In terms of licensing, we will have to get McNeel approval, as we will not be able to buy dozens of yearly licences, just running for a short while when in demand. So, we will see..
Our system doesn't currently implement a genetic algorithm (or any other solver / optimizer). But when we do implement it, it will not require any user skills that are not required to operate Galapagos. Certainly it will not require the user to divide the job by himself.
We have a central database that all instances have (network) access to, and we save output files which can later be accessed by slider values used to create that file.
Yes, it is possible to implement over remote internet connections and managed using web interfaces.

Amit

Permalink Reply by Daniel Kolling Andersen on June 2, 2015 at 7:57am