Taming the OOM killer
Posted Feb 5, 2009 22:49 UTC (Thu) by
tbird20d (subscriber, #1901)
In reply to:
Taming the OOM killer by brouhaha
Parent article:
Taming the OOM killer
I'm still baffled as to why this is an issue at all. IMNSHO, the ability to overcommit memory should never have been created in the first place. If you need more memory, buy more memory, or create a larger swap partition or file.
This is not an option in embedded devices - particularly those that
ship in the millions of units.
What user-space programs are allocating so much more memory than they actually need, anyhow?
This simple answer is, in a modern system, all of them.
Nearly all programs have a virtual memory footprint greater than they will actually use (at one time, or even during their entire lifetime). This is true even ignoring the fork-and-exec issue. For large systems, this discrepancy is easily ignored or worked around, but for low-resource systems, the difference becomes a major issue. The ability to overcommit memory is actually one of the reasons Linux is selected over traditional RTOSes for some embedded projects.
(
Log in to post comments)