I'm surprised the GPU is a win when the data is coming from GCS. The CPU still has to touch all the data, right? Or do you have some mechanism to keep warm data live in the GPUs?
As you say the issue is that you have an overall process to optimise from getting the data off slow GCS onto the nodes, shuffling it which often then writes it to a slow disk before the real processing even starts then writing back to a slow GCS.
Also, what do you think would be the best way to structure such a post?
But, here's a small bit of something perf-y: during large shuffles, I was able to increase overall job performance/efficiency by using external shuffles, even with times of ~5s median shuffle write for a couple hundred MB partitions (I hope I'm remembering this correctly, lol). This is not particularly great, but it did allow for cost-efficiently chewing through some rather large datasets without dealing with memory issues. There's also an awesome side benefit in that it allows us to use cheap spot workers in more scenarios.