Messages in this thread | | | Subject | Re: XFS and journalling filesystems | Date | Sun, 30 May 1999 00:12:15 -0600 | From | Larry McVoy <> |
| |
: It is possible by design on those Buses for any device acting as a master : to directly talk to another device acting as a target
This does not work out at all in practice. It sounds very simple (and is if all you want is to dd if=/dev/hda of=/dev/hdb) but starts to fall apart immediately. Consider all the different devices and how they provide the byte stream abstraction and you start to see that the OS is frequently involved in the process.
I've been involved, briefly, in the design of two such systems before I realize what a mistake it was. After I pulled away from the last one, I watched almost a million lines of code get writtten before they gave up, realizing they were reimplementing what the OS already did.
Limiting your I/O device<->device throughput to 1/2 memory bandwidth is not an unreasonable thing. And I don't say that lightly, I feel the need for speed everyday, but not at the expense of the OS...
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |