Return to BSD News archive
Xref: sserve comp.unix.admin:14269 comp.unix.bsd:12767 comp.unix.internals:6347 comp.unix.wizards:31243 Newsgroups: comp.unix.admin,comp.unix.bsd,comp.unix.internals,comp.unix.wizards Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!agate!howland.reston.ans.net!pipex!uknet!gdt!aber!fronta.aber.ac.uk!pcg From: pcg@aber.ac.uk (Piercarlo Grandi) Subject: Re: why is altering MINFREE discouraged ? In-Reply-To: vds7789@aw101.iasl.ca.boeing.com's message of Fri, 15 Oct 1993 21: 26:23 GMT Message-ID: <PCG.93Oct18191520@decb.aber.ac.uk> Sender: news@aber.ac.uk (USENET news service) Nntp-Posting-Host: decb.aber.ac.uk Reply-To: pcg@aber.ac.uk (Piercarlo Grandi) Organization: Prifysgol Cymru, Aberystwyth References: <1993Oct15.212623.14509@aw101.iasl.ca.boeing.com> Date: Mon, 18 Oct 1993 18:15:20 GMT Lines: 15 >>> On Fri, 15 Oct 1993 21:26:23 GMT, vds7789@aw101.iasl.ca.boeing.com >>> (Vincent D. Skahan) said: Vincent> I can't seem to find anything other than the statemen "setting Vincent> minfree below 10% can cause performance problems." The rationale is described in the original FFS paper, cited in another thread in thir newsgroup. Briefly: when looking for a cylinder group with free blocks to extend a file, the FFS chooses a cylinder group *at random*, for some rather good reasons. If the amount of free space is below a certain threshold, it must try several cylinder groups to find one with free blocks, and this is rather expensive. The threshold is around 10% for 8KB/1KB blocks and 5% for 4KB/512B blocks, according to the paper.