> ... on practically every single context switch ... How frequently do context switches happen? Can this be reduced, so that there is less disk thrashing? I.e.

> ... on practically every single context switch ... How frequently do context switches happen? Can this be reduced, so that there is less disk thrashing? I.e. whatever process needs lots of memory swapped in, just don't switch to it for up to maybe a few seconds, rather than milliseconds or whatever? This seems less bad than having the OOM killer just kill the thing.