I haven't thought about it much, by my take on this is that a) we should separate pagercache limitation and mapped memory limitation completely.
b) mapped memory limitation is a form of RSS control. We need that for at least three reasons: a) To implement RLIMIT_RSS (finally) b) For isolation of vserver-style instances c) CKRM memory controller any implementation should be suitable for all these applications. There's a VM minisummit in Ottawa July 16 (organised by Martin Bligh). I have asked that this RSS management problem be placed on the agenda. _______________________________________________ ckrm-tech mailing list https://lists.sourceforge.net/lists/listinfo/ckrm-tech
