@ed709062 Microsoft Research’s Dryad got productized (search keyword: “LINQ to HPC”) and made it all the way through three beta releases with numerous customers planning to build on top of it; it got killed when some guy in Azure promised that he could make Hadoop work better with a minimal amount of time spent on “optimizing” it. He couldn’t.

Too bad; “write computation DAGs in C# and have the runtime schedule them across large distributed datasets” is an unmatched feature even a decade later.