From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Brown Subject: Re: Triple-parity raid6 Date: Sat, 11 Jun 2011 20:05:32 +0200 Message-ID: References: <20110609114954.243e9e22@notabene.brown> <20110609220438.26336b27@notabene.brown> <87aadq5q1l.fsf@gmail.com> <4DF20C18.3030604@christoph-d.de> <20110611101312.GA3528@lazy.lzy> <20110611131801.GA2764@lazy.lzy> <4DF38424.1010500@gmail.com> <4DF3A27E.8080806@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <4DF3A27E.8080806@gmail.com> Sender: linux-raid-owner@vger.kernel.org To: linux-raid@vger.kernel.org List-Id: linux-raid.ids On 11/06/11 19:14, Joe Landman wrote: > A quick note of caution (and someone from Netapp, feel free to speak up). > > Netapp has a patent on triple parity raid (c.f. > http://www.freepatentsonline.com/7640484.html). A quick look over this, > suggests that the major innovation is the layout and computation which > they simplified in a particular manner. That is, I don't think their > patent covers triple parity RAID in general, but does cover their > implementation, and the diagonal parity with anti-diagonal parity > (effectively counter propagating or orthogonalized parity). > > I am not sure what this means from a coding sense, other than not to use > their techniques without a license to do so. If Netapp wants to grant > such a license, this would be good, but I suspect that it wouldn't be > quite as simple as this. > > Just a note so that we don't encounter problems. I think its very > possible to avoid their IP, as it would somewhat hard to claim ownership > of the Galois Field math behind RAID calculations. They can (and do) > claim a particular implementation and algorithm. > > [also not trying to open the patent on code wars here, just pointing out > the current situation ] > > I've read a little about diagonal parities - I can see some advantage in their simplicity, but I think that they are a poor choice for raid. Raid5+ already suffers from performance issues because you often have to read a whole stripe at a time just to change a few blocks - with diagonal parity, you'd have to read a whole 2-D set of stripes.