It's long been well known that RAID5 creats additional overhead (thanks to parity information) which makes it slower than say RAID0 (striping).
While following up something I was reading today (distributed.net) I came across this 3ware page, touting the performance of their recent SATA2 RAID contoller. While their main point is to compare with their previous SATA1 RAID controler (and the performance increase that they've managed), just look at the RAID5 vs RAID0 performance.
Read is only 13% behind, write is only 9% behind, and across 8 drives, you've only lost 12.5% space.
That's pretty hot in my book, that's a paltry 11.5% penalty (if you want to think of it that way) to achieve coverage against single drive failure... Imagine how the 12 and 16 drive versions do!
BTW they don't mention how much on-board cache they're using, I'm guessing the full 256MB.
No comments:
Post a Comment