On 07/22/2013 11:26 AM, Chen, Xiaoxi wrote: > > > 发自我的 iPhone > > 在 2013-7-23，0:21，"Gandalf Corvotempesta" <gandalf.corvotempe...@gmail.com> 写道： > >> 2013/7/22 Chen, Xiaoxi <xiaoxi.c...@intel.com>: >>> Imaging you have several writes have been flushed to journal and acked，but >>> not yet write to disk. Now the system crash by kernal panic or power >>> failure，you will lose your data in ram disk，thus lose data that assumed to >>> be successful written. >> >> The same apply in case of journal failure with data still on it. >> Imagine an SSD journal with 50GB of data. If SSD fails, all datas are lost. >> >> The only difference is that RAM is volatile and subjected to kernel >> panics or power failure (I only have dual power server) but actuall >> RAM is *MUCH* more reliable than SSD. I've never seen a single RAM >> module (server grade) failed from the latest 5-6 year. > RAM is physically much more reliable than ssd,but when taking kernel/power > failure into account, i would like to bet Ram disk is MUCH dangerous than ssd
I have not yet had the opportunity to try one, but something like the Marvell Dragonfly might be a very interesting option for servers with 24+ drives: https://origin-www.marvell.com/storage/dragonfly/nvram/ Mark > _______________________________________________ > ceph-users mailing list > ceph-users@lists.ceph.com > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > _______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
