From owner-freebsd-hackers Thu Aug 2 11:46:53 2001 Delivered-To: freebsd-hackers@freebsd.org Received: from InterJet.elischer.org (c421509-a.pinol1.sfba.home.com [24.7.86.9]) by hub.freebsd.org (Postfix) with ESMTP id 930A237B401 for ; Thu, 2 Aug 2001 11:46:50 -0700 (PDT) (envelope-from julian@elischer.org) Received: from localhost (localhost.elischer.org [127.0.0.1]) by InterJet.elischer.org (8.9.1a/8.9.1) with ESMTP id NAA47000; Thu, 2 Aug 2001 13:48:22 -0700 (PDT) Date: Thu, 2 Aug 2001 13:48:21 -0700 (PDT) From: Julian Elischer To: David Gilbert Cc: Freebsd-hackers@freebsd.org Subject: Re: _real_ ram disk needed ... In-Reply-To: <15209.36726.676011.413494@trooper.velocet.net> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG It might be possible to make a device that could acces > 4GB using the new paging methods available in PIII but it might require suspending th erest of the OS while you do it, and having a 4MB 'window' into which you COPY data to and from the extended space. (still it may be faster than disk) On Thu, 2 Aug 2001, David Gilbert wrote: > It would seem that the semantics of mfs and md differ signifcantly. > If I write data to mfs, it is only sent to swap if it absolutely must > --- I chalk this upto the allocations it uses operating like process > memory. If I write to md, the speed at which I write to md (swap > backed) is limited by the speed of the swap device. It would appear > that only a very limited amount of write buffering is going on. > > Now... I'm dealing with a very poorly designed _binary_ application. > It writes out very large sparse matrices to disk (as much as 6G, but > randomly between a few hundred meg and 6 gig). It is desireable for > the system to _not_ write the blocks in question to disk unless > absolutely necessary. > > Is there any way to make this request to the vm system? > > For something that fits in memory, the performance of mfs is many > dozens of times faster than md. > > Dave. > > -- > ============================================================================ > |David Gilbert, Velocet Communications. | Two things can only be | > |Mail: dgilbert@velocet.net | equal if and only if they | > |http://www.velocet.net/~dgilbert | are precisely opposite. | > =========================================================GLO================ > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-hackers" in the body of the message > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message