From mboxrd@z Thu Jan 1 00:00:00 1970 From: Loic Dachary Subject: Re: Checking of Reed-Solomon Vandermonde parameter combinations Date: Thu, 03 Jul 2014 12:30:26 +0200 Message-ID: <53B530C2.5020401@dachary.org> References: <53B52EC8.5030708@dachary.org> <3472A07E6605974CBC9BC573F1BC02E4AE747EDC@CERNXCHG44.cern.ch> Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="bdu56dnDcRxRcXS0Mb1FDGllALqB0tI1O" Return-path: Received: from mail2.dachary.org ([91.121.57.175]:36717 "EHLO smtp.dmail.dachary.org" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1750978AbaGCKac (ORCPT ); Thu, 3 Jul 2014 06:30:32 -0400 In-Reply-To: <3472A07E6605974CBC9BC573F1BC02E4AE747EDC@CERNXCHG44.cern.ch> Sender: ceph-devel-owner@vger.kernel.org List-ID: To: Andreas Joachim Peters Cc: Ceph Development This is an OpenPGP/MIME signed message (RFC 4880 and 3156) --bdu56dnDcRxRcXS0Mb1FDGllALqB0tI1O Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Thanks for the clarification, we're good then :-) On 03/07/2014 12:27, Andreas Joachim Peters wrote: > Hi Loic,=20 > as much as I understood for now it should not happen in Jerasure, but i= t will happen with the default Vandermonde matrix in the Intel library. T= he reason is the way these matrices are constructed. >=20 > But to verify this you have to make millions of iterations not only 1 b= ecause for large (k,m) there many failure scenario (chunk combinations) t= o test. >=20 > Cheers Andreas. >=20 > ________________________________________ > From: Loic Dachary [loic@dachary.org] > Sent: 03 July 2014 12:22 > To: Andreas Joachim Peters > Cc: Ceph Development > Subject: Checking of Reed-Solomon Vandermonde parameter combinations >=20 > Hi Andreas, >=20 > There are some combinations of K/M parameters for which Reed-Solomon Va= ndermonde may not be able to recover from the loss of M chunks. Reed-Solo= mon Cauchy does not suffer from this problem. (I'm paraphrasing a convers= ation we had in private, feel free to correct me if I'm wrong). >=20 > In the context of Ceph we are interested in a limited range of paramete= rs so I ran a test with: >=20 > https://github.com/ceph/ceph/blob/master/src/test/erasure-code/ceph_era= sure_code_benchmark.cc#L40 >=20 > for k in $(seq 2 50) ; do for m in $(seq 1 $k) ; do echo -n "k=3D$k m=3D= $m " ; ./ceph_erasure_code_benchmark --plugin jerasure --parame= ter directory=3D.libs --parameter technique=3Dreed_sol_van --pa= rameter k=3D$k --parameter m=3D$m --erasures $m --iterations 1 = --workload decode ; done ; done >=20 > and it does not throw an error. It basically take a range of parameters= K=3D2,M=3D2 up to K=3D50,M=3D50 and decode with M erasures for each of t= hem. Should some of them fail ? >=20 > Cheers > -- > Lo=EFc Dachary, Artisan Logiciel Libre >=20 --=20 Lo=EFc Dachary, Artisan Logiciel Libre --bdu56dnDcRxRcXS0Mb1FDGllALqB0tI1O Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iEYEARECAAYFAlO1MMIACgkQ8dLMyEl6F21SfACdHHWaLfMWuTXr2h5qtNGZ3Zy0 cosAoLGBN7NQNLlzGAWdAsdiB/WO3rmj =jAFH -----END PGP SIGNATURE----- --bdu56dnDcRxRcXS0Mb1FDGllALqB0tI1O--