On Mon, 2005-03-21 at 14:04 -0600, Aleksandar Milivojevic wrote: > Christofer C. Bell wrote: > > > Aleksander, can you please name some mainstream commerical Unix > > operating systems that are designed to be "rock stable in real world > > applications" using the criteria you've outlined? (Those criteria > > seeming to be kernel level process starvation protection that's > > "always on" and "intelligent"). > > I have to admit, none in Unix category that is absolutely perfect (out > of those I worked with). But there are some not far from there. For > example, OpenBSD can't be brought down by single process being swapped a > lot. The mmap/memset attack simply doesn't work on it. Things do get a > bit on a slowish side, however services are still responsive and I'm > perfectly able to login and kill offending process (just tested it on > ancient sparc station 5 with 64 meg of RAM and old and *slow* disks that > can hardly do 2 megs/sec). On Linux, no such luck, brings fast P4 > machine (with relatively fast disks capable of 20 megs/sec) to a halt. > > Solaris handles this kind of things more gracefully, as well as old > Digital's OSF/1 (alias Digital Unix, alias Compaq Tru64, alias whatever > HP calls it nowdays). Putting Aleksander on direct CC because I really want to know this. On the BSD boxes, if you do a fork bomb on the same ulimits does it crash(I'm assuming of course thet BSDs have ulimtis too)? I would think not. (per your explanation above) Hence, it's no wonder why I wonder why that is so for the BSD box. What is the difference between how LInux and how BSD handles forks? > -- Ow Mun Heng Gentoo/Linux on DELL D600 1.4Ghz 98% Microsoft(tm) Free!! Neuromancer 18:02:12 up 8:39, 6 users, load average: 0.78, 0.56, 0.50
