@_JacobTomlinson@TEGNicholasCode@xarray_dev Thanks for the suggestions! Rechunking is always going to be a bottleneck, but I'd be pleased if we could get to the point in Cubed where there is one intermediate store per rechunk operation in a workflow.
@al_merose@TEGNicholasCode Lots of interesting ideas! BTW there's a bi-weekly meeting to discuss these kind of things, that you may be interested in: https://t.co/ec1sZ7kj3H
@_JacobTomlinson@TEGNicholasCode@xarray_dev Also, I think there are some very significant optimization opportunities to get the number of writes to intermediate storage down to something much more reasonable. This is probably the biggest: https://t.co/Q72X0blVI8 (see red circle).
@_JacobTomlinson@TEGNicholasCode@xarray_dev Thanks Jacob! I agree that using memory/lower-latency intermediate storage is worth trying (https://t.co/mZn3iZuoSB). Do you know any cloud memory caches that are serverless? (The ones I've seen need a cluster.)
I want a future where an xarray workload can be executed like a SQL query, on a wide range of backends, who compete to provide the best performance in various cases.
https://t.co/FlCfM62LYh
One for the data engineers: I've been working on a new project called Cubed for serverless array processing. Tom Nicholas has written a superb blog post explaining how it works https://t.co/khkuKVNXKZ
🚨🚨🐍🐍PyConUK 2023 call for proposals is now open 🐍🐍🚨🚨
https://t.co/6nSdVwo6Hf
This call will close on Friday 30th June
We’re looking for main stage events, classroom events, lecture room events, young coders’ day events and any other ideas you may have.
Submit now!
New 'Night Science Podcast' episode! Mathematician extraordinaire @stevenstrogatz talks about how to ruthlessly simplify a problem to the point where - while it still seems impossible - it is down to its bare essentials. @ItaiYanai@MartinJLercher
https://t.co/NAF0QbdZdE
An exciting technical project we've been working on at https://t.co/DCdDfXPXxs allows us to build a fresh container image, and boot up a lot of containers running that image, on many different nodes, all in a couple of seconds. Some notes on how we do it: