"never implemented a proper deduplication system, so had duplicate copies of a lot of data for that reason."
That's what I took from the article. Very large amounts of data take time to dedupe, especially across 8 separate arrays of spinning rust. Deduping on flash all in one box and using the flash to assist the dedupe of the spinning rust makes it much fast and more manageable without hitting on the array performance.
Biting the hand that feeds IT
