James, On Sunday 06 March 2005 05:07, James Knott wrote:
One thing I've noticed, is that while there's been lots of talk, over the years, about minimum swap partition size, I don't recall seeing mention of a maximum size. For example, 32 bit CPUs have a maximum addressing range of 4 GB.
That is the maximum conceivable _virtual_ address space. Each process has its own, so if you wanted to make sure that no process was limited by available virtual memory (physical + swap), you'd need the maximum (which is less than 4 GB because of reserved areas of the virtual space) times the number of concurrently executing processes you want to run. This approach to swap space sizing would yield a grossly excessive amount of swap space, of course.
Since swap is the virtual memory area, does this mean that it's also limited to 4 GB? If so, does that mean that the 64 bit CPUs can have a swap partition(s) of up to 16 exabytes?
The same analysis applies to 64-bit architectures. If you're actually going to be running programs that use the larger virtual address space of a 64-bit architecture, you'd want to get more physical memory and possibly configure more swap space, too. But keep in mind, your system will be pretty much unusable whenever the working set size (the amount of memory system-wide that is being actively accessed) gets much bigger than the available (non-kernel) physical memory. When that happens, paging activity increases sharply while CPU utilization drops because processes are spending most of their time waiting for their virtual memory contents to be paged in.
tnx jk
Randall Schulz