search

akiradeveloper / dm-writeboost

Log-structured Caching for Linux
GNU General Public License v2.0
123 stars 19 forks source link

ioprio hinting to bypass cache (or writeback cache) #174

Closed marcin-github closed 7 years ago

marcin-github commented 7 years ago

Hi! I' dont know if you are aware of this thread http://www.spinics.net/lists/linux-bcache/msg04214.html . Would it be hard to add similar feature to dm-writeboost?

akiradeveloper commented 7 years ago

No.

Again, writeboost can't bypass writes. Bypassing writes would result to destructing the consistency when the caching device dies. It's not about implementation difficulty but correctness. I think bcache sacrifices the property that consistency is maintained after caching device corruption and it's only tradeoff. It's just I am going the different way.

It's also important for writeboost to go the other ways than other caching solutions. Otherwise, writeboost will be just crap. That writeboost can attract some attention is just because it's unlike bcache or dm-cache. Making bcache-like something again is completely meaningless.

Only features that naturally adaptable to writeboost's concept and architecture will be implemented.

But yeah, thanks for notifying me the bcache's feature that I didn't know.

marcin-github commented 7 years ago

Ah, yeah, it was obvious, writeboost keeps order of writed blocks. Thanks for quick reply!

akiradeveloper commented 7 years ago

writeboost keeps order of writed blocks

exactly