> main argument for using mmap() instead of standard I/O is the fact > that reading entails just touching memory - in the case of the memory > being resident, you just read it - you don't even take a page fault > (so no overhead in entering the kernel and doing a semi-context > switch).
Oh and in the case of Java/Cassandra, as Jonathan clued me in on earlier, there is also the issue that byte[] arrays are mandated to be zeroed when allocated which causes overhead typically because there has to be a loop[1] somewhere writing a bunch of zeroes in, that you're then just going to replace immediately. Mapping a file has no such implications as long as you read directly from the underlying direct ByteBuffer. [1] Not *necessarily*; a JVM could theoretically do byte[] allocations in such a way that it already knows the contents is zeroed, but it would be highly dependent on the GC/memory management technique used by the JVM whether this is practical. (It just occurred to me that Azul should get this for 'free' in their GC. Wonder if that's true.) -- / Peter Schuller
