From owner-freebsd-questions Wed May 29 20: 3:51 2002 Delivered-To: freebsd-questions@freebsd.org Received: from altair.mukappabeta.net (altair.mukappabeta.net [194.145.150.157]) by hub.freebsd.org (Postfix) with ESMTP id 9175737B404 for ; Wed, 29 May 2002 20:03:45 -0700 (PDT) Received: by altair.mukappabeta.net (Postfix, from userid 1001) id 868AD6BF; Thu, 30 May 2002 05:04:13 +0200 (CEST) Date: Thu, 30 May 2002 05:04:13 +0200 From: Matthias Buelow To: Peter Leftwich Cc: Dan Nelson , FreeBSD LIST Subject: Re: mv vs. tar [overlapping] Message-ID: <20020530030413.GA1063@altair.mukappabeta.net> References: <20020530022609.GA956@altair.mukappabeta.net> <20020529223005.O79882-100000@earl-grey.cloud9.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20020529223005.O79882-100000@earl-grey.cloud9.net> User-Agent: Mutt/1.3.28i Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Peter Leftwich writes: >Actually, it's an interesting point... I think what Dan means is that read >and write processes run concurrently (have you ever run multiple HTTP >requests at once? it's a lot like that I suppose). So it would seem >tarring is quickest when moving a directory within the same filesystem! It of course _may be_ that this method exploits a certain advantage in how blocks are read/written from/to the same disk (or two disks.) It certainly adds some (fixed) computational overhead, which would be measurable on something like an old VAX-based machine but which should basically have no reasonable effect on today's modern processors. However. Without (even empirical) proof, I somehow doubt that there is much difference in I/O-performance between reading and writing blocks in one process and doing the same in two different ones; I mean, the disk doesn't seek faster all of a sudden because of that, and if you read n blocks in process 1, and write them in process 2, or do the same in one process, it should be quite the same. There might be a point when the files are mostly in the buffer cache and the cache can be accessed simultaneously without interference by two processes running each on a different processor on a SMP system but then again, that might be a bit far-fetched. --mkb To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message