Improve Parity performance in Storage Spaces
Give options or workarounds to improve parity performance in Storage Spaces.
It is not mentioned but I think everyone will agree that it is the write performance that needs a huge improvement. The read performance is acceptable to me. Still a problem in ws2019
Something needs to be done here. Parity would be so useful if it wasn't so painfully slow. Using 3 4TB HDDs with 120MB/s write performance each, I'm still getting only 30MB/s on average (~65MB/s max) for long sequential writes. Already tried different caching modes and sector size is already at 4k (drives are 512e, so support 4k)
Seek out existing hardware offload capabilities and use them. Be it the classical RAID6 engine or a network card, or even a special CPU, graphics card. Enable larger Galois fields.
Mark Fowler commented
And give us the option to modify the write cache, updating the cache algorithm to more effectively utilise much larger cache sizes.
(Also, if a write is bigger than a stripe full of data, then there is no need to read the previous contents to calculate parity, right?)