> There is an other option.
> The caching can still be done by the kernel buffers, but we could change
> them a bit so there is a new style of write() which does the following:
> start an asynchronous write of the data.
The whole point of raw devices access - is that its wasted memory to
buffer it. Any and all useful buffering of the data will be done by
the application. You will configure the application to use as much
memory as you think is necessary, for example, I've setup databases
and told them to use 350MB of 512MB machines. That way, the database
hopefully _never_ pollution memory outside of this specified high
water mark, and because the database knows with some detail how much
cache is in use, and how its used, it can may effective smart
decisions which helpfully will increase its performance on the whole.
Any additional copying and caching is wasteful. Its that simple.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to email@example.com
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html