Date: Tue, 11 Apr 2006 12:22:33 +1000 From: Tony Maher <anthony.maher@uts.edu.au> To: Pete Slagle <freebsd-stable@voidcaptain.com> Cc: freebsd-stable@freebsd.org Subject: Re: Maximum Swapsize Message-ID: <443B12E9.2040107@uts.edu.au> In-Reply-To: <443B0A51.8040206@voidcaptain.com> References: <1dbad3150604100913hff9fc4dsb125ea541675f992@mail.gmail.com> <20060410161713.GA48094@xor.obsecurity.org> <200604111048.09905.doconnor@gsoft.com.au> <443B0A51.8040206@voidcaptain.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Pete Slagle wrote: > Daniel O'Connor wrote: > >> The old "swap size = 2x RAM" rule is no longer applicable unless you >> have a very special application. > > This "rule" always seemed counterintuitive to me anyway. > > When you have very limited physical RAM you need a lot of swap space. > When you have more than enough RAM you don't need any swap space at all. > For a given set of applications, as RAM increases you need less swap > space, not more. And vice versa. Provided the maximum "working set" of processes fits into RAM, you have sufficient RAM. Seldom used processes can be swapped out with minimal impact on the system. So as well as the "very special aplication" exception, the workload patterns (over a day) may allow for reduced RAM and utilize swap instead. In which case swap size should be sized to match. Maybe not important for a single machine but for multiple machines the cost of RAM memory adds up. -- tonym
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?443B12E9.2040107>