Date: Wed, 15 Feb 2012 17:44:18 +0200 From: "Pavlo" <devgs@ukr.net> To: "George Kontostanos" <gkontos.mail@gmail.com> Cc: freebsd-fs@freebsd.org Subject: Re: ZFS and mem management Message-ID: <4633.1329320658.10597993752278204416@ffe16.ukr.net> In-Reply-To: <CA%2BdUSyqRd9DFhhAtQqU%2BVdmUz_kxqo=jSApoGOUd1Q9StrCkxA@mail.gmail.com> References: <CA%2BdUSyqRd9DFhhAtQqU%2BVdmUz_kxqo=jSApoGOUd1Q9StrCkxA@mail.gmail.com> <41082.1329320114.6955040073494560768@ffe1.ukr.net> <CA%2BdUSyp1p07ERXaL-zLG4ibGB45c7VGBNH3ALL4ZXbXUbeEDiA@mail.gmail.com> <70229.1329318412.9319724204137054208@ffe16.ukr.net>
next in thread | previous in thread | raw e-mail | index | archive | help
--- Original message --- From: "George Kontostanos" <gkontos.mail@gmail.com> To: "Pavlo" <devgs@ukr.net> Date: 15 February 2012, 17:39:46 Subject: Re: ZFS and mem management > 2012/2/15 Pavlo <devgs@ukr.net>: > > > 2012/2/15 Pavlo <devgs@ukr.net>: >> >> >> >>>On 15/02/2012 13:39, Pavlo wrote: >>>> >> >>>> >> >> Unfortunately we can't afford disabling prefetch. It is too much >>>> >> >> of an>> overhead.>> >> Also I made some tests. I have process that maps file >>>> >> >> using mmap() and>> writes or reads first byte of each page of mapped file >>>> >> >> with some data.> >>>Note that ZFS is designed so that it interacts somewhat badly with >>>mmap() and other kernel services which rely on coherency between VM and >>>IO such as sendfile(). At the very best, you will have two in-kernel >>>copies of all data buffers used with such interfaces, but there have >>>been sporadic reports that there are other bugs with it. >>> >>>If you have a test server, I'd recommend you do the same test on UFS for >>>comparison. >> >> Was going to try this... Thanks for reply. >> _______________________________________________ >> freebsd-fs@freebsd.org mailing list >> http://lists.freebsd.org/mailman/listinfo/freebsd-fs>>; To unsubscribe, send any mail to "freebsd-fs-unsubscribe@freebsd.org" > > Why do you think that disabling prefetch is an overhead? > > > -- > George Kontostanos > Aicom telecoms ltd > http://www.aisecure.net>; > > > Well... not me though. System administrator >_> . I suppose because we have > a big IO traffic. Not for a highly random I/O environment. -- George Kontostanos Aicom telecoms ltdhttp://www.aisecure.net > That's is a reason we don't disable. Our reads and writes mostly contiguous. And that problem arises on machines with low RAM (8Gb and less). Also for obvious reasons swap is disabled.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4633.1329320658.10597993752278204416>