On 11/14/05, Eric Lowe <[EMAIL PROTECTED]> wrote:
While it sounds interesting from an academic point of view, I wonder how relevant this work (and any work related to paging) is when memory can be purchased for under $200 a gigabyte. This is a different perspective
Memory is cheap, but the system to put a reasonable amount into is quite expensive. According to store.sun.com: T2000 8 cores @ 1.2 GHz, 16 GB RAM: $21,495 T2000 8 cores @ 1.2 GHz, 32 GB RAM: $32,395 T2000 8 cores @ 1.4 GHz, 64 GB RAM: $84,995 If my app needs 32 GB of RAM, it looks to me as though the RAM greater than 16 GB is costing me $681 / GB. If I need an obscene amount of RAM, then the memory about 16 GB is about $1300 (allowing $733 extra cost for 0.2 GHz CPU bump). This idea seems very attractive to me for workloads that need memory to be faster than disk, but not necessarily top performance.
paging candidate determination very costly). ZFS is very fast (being copy- on-write it does sequential I/Os whereas swapfs does random disk access), and it supports compression. ZFS is already slated to become the new replacement to swapfs in the near future.
If free() could trigger the memory to be zero'd, then those pages would compress very well in a ZFS swap device. Presumably java's garbage collector could do the same zeroing, and make an explicit call to the OS to say "put this page at the top of the to-be-paged-out list." Mike -- Mike Gerdts http://mgerdts.blogspot.com/ _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
