On Monday 07 March 2005 17:12, James Knott wrote:
Bruce Marshall wrote:
On Sunday 06 March 2005 09:17 pm, John Sowden wrote:
Actually, I thought this was a great question that only one person came close to answering. Is there a defined set of tradeoffs when expanding the swap size, in relation to RAM, taking into consideration CPU speed and possibly other variables that I am not aware of. If, instead of using the 2X rule for sizing a swap partition, we went to a 4X or 6X size, would the paging overhead not justify the expanded swap size, in terms of increasing performance?
The best swap area you can have is one that doesn't get used. The size of the swap area should be immaterial as long as it is large enough to handle all of the swap. But if you are swapping a lot, you should look elsewhere for performance gains. (as in increasing your RAM...)
One situation I had in mind, was processing huge data files, so large that you wouldn't want to pay for all that RAM. Is processing it in memory space easier than doing the file I/O, to process it in chunks?
USually applications that access large files use memory-mapping via the mmap() call. The mmap() call converts an open file into a pseudo memory area. Accessing this memory area causes the kernel to do the necessary file I/O transparently. SGI had a whole host of video processing applications that used this technique. Regards Paul -- Paul Hewlett (Linux #359543) Email:`echo az.oc.evitcaten@ttelweh | rev` Tel: +27 21 852 8812 Cel: +27 72 719 2725 Fax: +27 86 672 0563 --