I am just looking for more accurate information around RAID-DP write penalty. I know that most folks at NetApp or NetApp consultants elsewhere would immediately say - 1 to 2 maximum. That's b'cos WAFL turns incoming radom IOs into sequential full stripes. That's understood, but is that an average figure of both during low storage utilization & high storage utilization ?
I would be really happy if someone can share more information around this, during: 1. Low utilization of storage 2. High utilization of storage
1. During low utiliztion of storage - WAFL collects logically random writes and turns them into physical full sequential writes, as a result of which, it calculates the parity in the meomory on the fly, that means write penalty of maximum 2.
2. During High utilization of storage - What if my aggregate is 95%, and I have fragmented spaces, free full stripes are out of question, club this with back-2-back CPs. WAFL have to read the 'on disk blocks' to re-calculate the Parity and then write the new stripe. Does this mean, during times like that, we could see write IO penalty going up to 4.