On Apr 6, 2009, at 5:28 PM, Eugene wrote: > ------------------------- > SNIP------------------------------------------ > There are days I wish OS X was more aggressive towards > consolidating swap. > > > Eugene… I agree. That also assumes a few things that are not apparent (at least from an outsiders view). One of the things swapping does is do BURST mode (that is many contagious frames that are needed to be swapped out and in smaller systems that is not always easy to tell (from an OS perspective that is) To run a robust paging system you need a lot more information than "should we swap it". You need when was it last used and *IF* it was altered. If it was not altered then there is no reason to page it out, as if the page is needed again it can just be recalled from where it was the last time. There are also items that the OS needs to give to the paging system like is there a paging shortage (Y/N) what tasks are waiting to be paged/swapped in. How much "free" (unallocated) storage is there. Is there a lot of demand paging (task(s) are waiting for a page in to take place). Then there is the time when real storage is short and you must put an active task to sleep so you can either swap of page out the task. I do not know a lot about how OS X and how I/O takes place on it. Can you chain say ten I/O's together or is it one page per page in/out per operation and here is the more interesting questions what size is each page is it 1K,4K,16K,32K,64K ? Maximizing the decision to swap in/out and or page(s) in/out is really almost a mathematicians dream. At least one OS does this and more. But in any case since MAC does not externalize the information so we are guessing here. Ed