From owner-freebsd-hackers Mon May 28 9:54:45 2001 Delivered-To: freebsd-hackers@freebsd.org Received: from host213-123-131-91.btopenworld.com (host213-123-131-91.btopenworld.com [213.123.131.91]) by hub.freebsd.org (Postfix) with ESMTP id 7EEF637B422 for ; Mon, 28 May 2001 09:54:38 -0700 (PDT) (envelope-from dominic@host213-123-131-91.btopenworld.com) Received: (from dominic@localhost) by host213-123-131-91.btopenworld.com (8.11.3/8.11.3) id f4SGsOT08405; Mon, 28 May 2001 17:54:24 +0100 (BST) (envelope-from dominic) Date: Mon, 28 May 2001 17:54:24 +0100 From: Dominic Marks To: "E.B. Dreger" Cc: freebsd-hackers@freebsd.org Subject: Re: brainstorm: "intermediate" disk caching Message-ID: <20010528175424.A1637@host213-123-131-91.btopenworld.> References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: ; from eddy+public+spam@noc.everquick.net on Mon, May 28, 2001 at 04:31:17PM +0000 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 Hi, (I'm geussing the 'public+spam' bit is standard removal stuff) On Mon, May 28, 2001 at 04:31:17PM +0000, E.B. Dreger wrote: > > array. Write performance is inherently disappointing -- which may or may In my opinion this is the same as how MFS when used without limitation can also be a bad thing. If you don't allocate a fixed amount then you risk having the cache expand so that it impedes the server's ability to function in a normal way. Equally if you set the cache too low then you either don't get a true benefit, or you can even end up in a situtation where the small buffer slows the ability of the server to operate. Another issue would be the process of moving the cache onto disc. This would require either a caching daemon process which you'd most probably put in the kernel (although I'm not fully qualified so say either way) which could end up being so busy moving cache data for example if your cache size was very small that it would starve CPU time and resources and again its has a negative effect upon the overall performance. Lastly it seems to me that the worst reason is that to cache RAID-5 would be to undo the very point of using it. It is designed for reliable data storage however caching it in memory seems to be the opposite of that principle as it could be wiped instantly with a power outage. In the worst case it could be that the cache was full at the time and without updates commited to permanent storage. I've assumed you ment memory when you talked of caching instead of on disc caching. The idea of perhaps caching writes onto a RAID-0 system and then transferring them is possible. But it sounds to that such a system would be hard to setup. > Eddy Just a few thoughts, I could be wildly out of focus Dominic To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message