From mboxrd@z Thu Jan 1 00:00:00 1970 From: Wido den Hollander Subject: Re: ceph data store size Date: Wed, 01 Aug 2012 15:33:21 +0200 Message-ID: <50193021.7070009@widodh.nl> References: <5018E379.4030208@lanning.cc> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from smtp01.mail.pcextreme.nl ([109.72.87.137]:55708 "EHLO smtp01.mail.pcextreme.nl" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753493Ab2HANdY (ORCPT ); Wed, 1 Aug 2012 09:33:24 -0400 In-Reply-To: <5018E379.4030208@lanning.cc> Sender: ceph-devel-owner@vger.kernel.org List-ID: To: lanning@lanning.cc Cc: ceph development On 08/01/2012 10:06 AM, Robert Hajime Lanning wrote: > Anyone using ceph on multi petabyte data stores? I don't think so > For example, a cluster of 12 systems with a combined storage of 2.3P. > With 12 systems, you mean 12 servers with a lot of disks? That's roughly 200TB per server. On that scale I don't think Ceph would work, it would mean you would need a HUGE amount of CPU and memory on those boxes to run the OSD's. If you want to scale to 2.3P you should be looking in the direction of hunderds of small machines all storing a couple of TB. Wido