Would it be possible to run this locally in computers/laptops with nvidia cuda drivers+docker?

[juankaamez]

Would it be possible to use this to run nodes of a cluster of computers which would be normal computers/laptops to create a beowulf-like infraestructure approach?
How would in that case be the database to fold proteins be distributed, meaning, would it be needed to download the database in each node? Would it be enough to download it in a distributed-database-like form?

[Gy-Lu]

Hi, thanks for your issue.
And the answer to your question is yes. It is possible to run it on normal computers/laptops with Nvidia GPU.
However, if you want to distribute saving your database, it should be organized carefully and the preprocessing code should be modified. Now FastFold has not offered such a feature yet.

[juankaamez]

Thanks a lot for your answer Gy-Lu!
I was wondering, if that were an option to be added (it might really be interesting to be able to switch from supercomputing or beowulf nodes creating a cluster), which distributed database model/engine would you suggest and how could we modify that preprocessing code?
It would allow to have nodes collaborating from all over the world, run locally, contribute with other supercomputing nodes in a heterogenous platform.
What do you think?

[Gy-Lu]

Hi, it sounds like a really interesting idea.
However, I am not familiar with DB, and I can not give you a suggestion.