Hi Christo, This can be a problem, depending on the size of the SGA and the amount of memory. There is a process that is responsible for reconstructing large pages from smaller ones but there are situations where it either can't work fast enough or can't relocate the necessary pages to do so. To make a single 4M page, the right 512 8K pages have to all be available, since the memory addresses have to be contiguous. Copying files tends to use a lot of filesystem buffers, which are usually 8K and so may break up many 4M pages. Those then have to be coalesced to provide 4M pages to the application afterwards. There are some workarounds that can help that happen and later versions of Solaris 10 do a better job of this, but it still can be an issue. Regards, Dave Miller Christo Kutrovsky wrote, On 10/09/08 16:05:
However the problem with Solaris is if you have the database down while you do ...say file copy operations, then the memory gets fragmented and when you start it up later it's using smaller pages / less of the "large" pages. However, this is not an issue on Solaris at all due to ISM.