On Sun, Apr 6, 2008 at 5:54 PM, Sam Clemens
Neil wrote:
If I am correct and swap space is striped over the different drives it is placed upon then it should matter more how many drives you have with swap spaces on it.
Swap space is striped ONLY if you set it up to be so. If two swap spaces have different priority levels, one swap space will be exhausted before the system starts to use the 2nd one.
Hmm, what should I do to make the swap space striped? Is there a command in the installer to do so?
True: the access time of old harddisks is
worse, but then again: the swap space will probably be used to store the bigger files anyways,
Swapping is not used to "store files"... it's used to page out memory pages so that the CPU can use a larger address space than the amount of physical memory installed (or more specifically: virtual address space = physical mem + swap space)
[This ignores annoying things like memory-mapped I/O, etc.]
True, storing is not the correct word. I thought Linux was able to see wether a file was big and swap it out completely. If it can't than that is to bad for my theorie.If you didn't see that it was a theorie then I should be clearer in that inthe future. Maybe some fake HTML like tags. I do know swapping is used to create a virtual memory to extend the RAM in times of great need.
since Linux is smart in handeling swap
space.
smart, yes. magic: no
The access time of a harddisk is always a order of magnitude
longer than the acces time of RAM, even with the newest (non solid state) disks, so IF the designers of the daemons that handle the swap space of Linux are as smart as I believe they are then (when swapspace is needed) the largest files will have a larger chance of getting in the swapspace (although there most likely are strong arguments why some files shoudn't be swapped if possible).
But swapping isn't done by file, it's done on a page by page basis. You can have some of a file in memory, some of it swapped out, and some of it not even read yet.
And pages to be swapped are evaluated on an individual basis (what's the current page size now...4k? 16k? 54k?)
Google for "demand-paged-virtual-memory", and take a look at these: http://en.wikipedia.org/wiki/Paging http://en.wikipedia.org/wiki/Page_replacement_algorithm http://en.wikipedia.org/wiki/Demand_paging
Note that Windows has historically NOT used "demand paging"...instead, what they call "aggressive paging" (anticipatory paging also known as prepaging) which in theory works great, but in practice, results in lots of paging activity which is never needed -- it seems to be one of the primary reasons why Windows is so pathatically non-responsive so often -- instead of just paging out what needs to be paged out, and then getting with running your program ... n ooooooooo, it decides to page out more memory, so that it can pre-load code within a branch that never gets executed. [Great plan you got there, Redmondites].
So: the resulting speed of your swap space is probably in lesser extend a result of the actual speed of the harddisk and to greater extend a result of the number of harddisks in swap.
I would have used my new drives for swap space, but you started a train of thougts causing me to reconsider that.Hmm, if I simply use the tower of 1 to 5 G 5400 rpm harddisks to create swap space that might save me from a crash in the future.
If your swap drive dies, you're still going to crash.
But if there's nothing else on the disk, at least you're not losing filesystems, too.
And that is what I need. I can survive a PC crash, but a harddisk crash would cost me money, so I wold like it to be as late as possible. Thanks for the info Neil -- There are two kinds of people: 1. People who start their arrays with 1. 1. People who start their arrays with 0. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org