The kernel uses 2 buffers, the buffer cache and the page cache, I have not seen the term read-ahead used with that. On the other hand, the settings mentioned in the earlier emails is from the IDE\ATA commandset and are finally set through the kernel ide driver, drivers/ide/ide-disk.c. Regards Amit Zou Min wrote: > Thanks for the pointer! > Actually, I am interested in disabling kernel readahead in order to do > some experiment. > > So, is there any fast way to do that? > > On Fri, Apr 12, 2002 at 02:46:47AM -0700, Sharath wrote: > >>guys.. guys.. do not mistake the kernel read ahead >>with the disk's... >> >>the harddisk contains a buffer (256k in mine and 1 mb >>in the newer models).. when a read request is issued >>to the disk, the disks logic decides which sectors to >>'buffer'... this data is stored in the HDs internal >>buffer.. and not transferred to the system's memory.. >>the disk's read ahead and is not controlled but the >>OS. >> >>the kernel has its own buffers to store IO. The kernel >>would be left to decide which sectors to read into >>these regions. If the process is IO bound and is >>reading a file, the kernel can predict the next read >>and 'prepare' for it (look into inode entries and >>determine where the next data recides) before the >>actual read call occurs. This is the kernels >>readahead. >> >>Ideally, they shouldn't be disabled. I have got some >>of the best performance with full utilization of my >>disk capabilities.. >> >>but don't have a clue on which algorithmn the kernel >>uses.. but this is twisting the matters a little.. >> >>but the bottom line is: hdparam is different from what >>the kernel does.. DO NOT mistake them > > -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
