Open stevengj opened 11 years ago
This may be less valuable, since much of the parallel APIs use apply
requests, rather than execute
requests, which use Python-specific serialization of objects (pickle), rather than the abstract JSON-able message spec. Execute requests, etc. do work, of course, as will things like the %%px
magic which uses execute. But it would still be interesting.
I was thinking that the cluster thing could just be used to launch or connect to Julia kernels, with the communication being done in the Julian way. But maybe I'm misunderstanding the IPython cluster facilities completely?
This may be a fruitful area for cross-language/platform abstraction since Julia already has its own mechanisms for spinning up clusters and doing distributed computing.
Ah - there should be nothing Python-specific about the startup / routing, just the client API.
Is it worth revisiting this? Jupyter still has a Clusters tab (which I know nothing about).
Right now the Notebook interface has a "Cluster" tab that I've been ignoring. But it would be good if someone could look into IPython's cluster machinery and figure out how to integrate it with Julia's parallel computing facilities.