If that’s the case then your benchmark is missing a peak memory consumption metric and the exact code used for it. It’s not clear for me at least based on the article.

Is Deephaven similar to other lazy evaluation libraries?