Date: Sun, 24 Oct 2004 00:17:18 +0100 (BST) From: Robert Watson <rwatson@freebsd.org> To: Igor Pokrovsky <ip@doom.homeunix.org> Cc: hackers@freebsd.org Subject: Re: Relative performance of swap-backed MFS vs. regular UFS? Message-ID: <Pine.NEB.3.96L.1041024001523.72477N-100000@fledge.watson.org> In-Reply-To: <20041023182420.GA19119@doom.homeunix.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 23 Oct 2004, Igor Pokrovsky wrote: > On Sat, Oct 23, 2004 at 01:12:46PM -0500, Ryan Sommers wrote: > > > > > >You can also use md(4). In my case I use it for /tmp. > > > > > MFS is the same thing as md(4). mfs = Memory File System, md = Memory > > Disk. Difference is only in the name. > > I thought mfs is allocated from virtual memory, while md - directly from > RAM. Am I wrong? When you use mdconfig(8) to configure the md device, you get to specify where you want the storage to come from. The possibilities include malloc (kernel address space + locked memory), but also vnode (file) and swap (pageable memory). Except during early netboot where swap is not available, I would recommend against using malloc-backed disks. For small disks it's OK, but as they get large they chew into kernel memory and address space, and so make it much easier for the system to run out of kernel address space. Robert N M Watson FreeBSD Core Team, TrustedBSD Projects robert@fledge.watson.org Principal Research Scientist, McAfee Research
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.NEB.3.96L.1041024001523.72477N-100000>