Sean Liu wrote:
Now - let's take a look from another angle: a. When people take a look at vmstat output, they don't expect an exact number of bytes of free memory, they need a rough number however. So some reasonably accurate number would be good enough.
good enough may be reasonably achievable, just subtract arcstats:c_min
b. For arguement sake, can we examine that "under extreme memory pressure, can kernel return (almost) all ZFS cache" b.1 Yes it can. Then all ZFS cache can be thought of free b.2 No, some of the key structure/metadata/whatever of N bytes must stay in memory. We should've known them when we allocated them. So cache size - N bytes counted to free memory b.3 No, but as a rule of thumb, X% of ZFS cache are returnable. Then cachesize * N% can be counted to free memory b.4 No, we have no clue at all. Let people stay in dark.
We lost the clue with the invention of demand paged virtual memory 30 years ago. And now you want to add layers of virtualization to get a better clue? Methinks you are asking the wrong question. -- richard _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
