I can't find in the article if those were sequential 1.5PB writes or random small writes (i.e. < 4KB). If the later case, this article should be flagged.