Very interesting. I've had something like this on my todo list for a while, and one feature I would love to try is a distributed heap using something like Datomic that you would do large allocations on so that you can care even less about where data processing computations run.
Yes, that's generally the kinds of things I'd expect to try. I have a number of chapel examples doing hash partitioning to locales. This makes it easy to have distributed indices.
Super interesting. I am currently building a distributed machine/deep learning framework that runs in node.js and the browser. Will definitely take a closer look at your project.
4 comments
[ 3.0 ms ] story [ 16.2 ms ] thread