From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from [82.179.117.26] (helo=shelob.oktetlabs.ru) by canuck.infradead.org with esmtps (Exim 4.62 #1 (Red Hat Linux)) id 1G65zP-00056z-Of for linux-mtd@lists.infradead.org; Thu, 27 Jul 2006 09:33:27 -0400 Message-ID: <44C8C07D.2050909@yandex.ru> Date: Thu, 27 Jul 2006 17:32:45 +0400 From: "Artem B. Bityutskiy" MIME-Version: 1.0 To: falls huang Subject: Re: question: the performance of jffs2 on UBI References: <49eab5c80607270530j469f2585w45fbe1d6449c2921@mail.gmail.com> <44C8BD55.3040604@yandex.ru> In-Reply-To: <44C8BD55.3040604@yandex.ru> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Cc: linux-mtd@lists.infradead.org List-Id: Linux MTD discussion mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Artem B. Bityutskiy wrote: > falls huang wrote: > >> The UBI and jffs2 both provide wear leveling. Will the repeated >> wear-leveling reduce the performance of system ? > > I guess jffs2 ought to be slower on UBI, but little. My *very* rough > test on mtdram device showed that JFFS2 over UBI is about 5% slower. But > more accurate testing should be done. And note: of course, JFFS2 was not designed for UBI, so it does not use UBI advantages. But still, it makes a lot of sense to use JFFS2 over UBI because in this case you have much freedom in partitioning your flash as you like. I'm planning to return to JFFS3 research/development some day and design JFFS2 to work on top of UBI devices. The alternative way to is to create a good block device layer over UBI, and then use conventional file systems (N.B. the mtdblock driver is *not* a good block device layer). -- Best Regards, Artem B. Bityutskiy, St.-Petersburg, Russia.