From mboxrd@z Thu Jan 1 00:00:00 1970 From: Marcel Lauhoff Subject: Re: OSD-Based Object Stubs Date: Thu, 28 May 2015 11:45:12 +0200 Message-ID: <877frtc9xz.fsf@irq0.org> References: <87a8wqct38.fsf@irq0.org> Mime-Version: 1.0 Content-Type: text/plain Return-path: Received: from cpu0.net ([37.187.2.217]:37549 "EHLO cpu0.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751548AbbE1JpK (ORCPT ); Thu, 28 May 2015 05:45:10 -0400 Received: from janelane.irq0.org (x5d82518d.dyn.telefonica.de [93.130.81.141]) by ks3366964.kimsufi.com (Postfix) with ESMTPSA id 307878029A for ; Thu, 28 May 2015 09:45:08 +0000 (UTC) In-reply-to: Sender: ceph-devel-owner@vger.kernel.org List-ID: To: Robert LeBlanc Cc: ceph-devel Robert LeBlanc writes: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA256 > > At first I thought this was to allow the OSDs to stub the location of > the real data after a CRUSH map change so that it didn't have to > relocate the data right away (or at all) and reduce the number of map > changes. Then I realized it was for cold storage tiering. Is the idea > to be able to move data off to near-line storage like tape? Yes, cold storage. Basically, let the OSD move data elsewhere like tape or some archive system. ~marcel -- Marcel Lauhoff Mail/XMPP: ml@irq0.org http://irq0.org