Suggestion - Standard HD Mode
Posted: Tue Mar 01, 2011 8:09 am
I know support mentioned that FC is optimized for SSDs in order to save writes to the drive. SSDs also have high burst capacity and really fast randoms. Standard mechanical drives however do not.
If possible it would be nice if FC reordered the data on it's own for write requests (sorta like NCQ or TCQ) only with a much bigger cache and array of data at it's disposal. This would definitely help with write times on a mechanical hard drive. It would also be nice if the data could be more streamlined, so if there is a burst of writes it's cached to FC and slowly trickeled out to cause as little of a impact as possible on HD performance (of course this depends on how big the write cache is). The same could be applied to reads in its own way, sorta like a vsync for hard drives.
I'm suggesting this as a completely different mode of operation that is selected by the user. I don't know if anything like this is already implemented. I'm using FC and I also have a three disk mechanical raid array I use so the current design of FC might not fit all users particular scenarios.
If possible it would be nice if FC reordered the data on it's own for write requests (sorta like NCQ or TCQ) only with a much bigger cache and array of data at it's disposal. This would definitely help with write times on a mechanical hard drive. It would also be nice if the data could be more streamlined, so if there is a burst of writes it's cached to FC and slowly trickeled out to cause as little of a impact as possible on HD performance (of course this depends on how big the write cache is). The same could be applied to reads in its own way, sorta like a vsync for hard drives.
I'm suggesting this as a completely different mode of operation that is selected by the user. I don't know if anything like this is already implemented. I'm using FC and I also have a three disk mechanical raid array I use so the current design of FC might not fit all users particular scenarios.
