I hardly can imagine situations when the OOM killer needs to be tuned so much. On my systems (notebooks/workstations/servers), it was always a user process whic

I hardly can imagine situations when the OOM killer needs to be tuned so much. On my systems (notebooks/workstations/servers), it was always a user process which allocated much more memory then available. In such situations, OOM killer works perfectly, and on systems without swap, it is a questions of seconds when the system is back to stable. (With swap, it takes minutes to fill the swap before OOM killer activates) On a productive and reliable system, I would rather invest in more RAM or load balancing then putting my time in fine tuning OOM killer...