It is quite common for RAID arrays and systems like Thumper to be sold with a complete set of drives. However, many sites likely start off with a relatively small amount of data and expect to grow into the space. The full storage cost is paid when the system is purchased.
While there is the option to map as many drives as possible into raidz or raidz2 VDEVs (e.g. set in stone) but there another option to consider which may offer more initial benefits while preserving future flexibility. It does not result in as much maximum total storage space. It idea is to initially configure the system using triple mirrors. triple mirrors offer more data reliability than even raidz2 and they offer considerably more read performance. With this approach all drives are exercised with live data, with more of the media being initially exercised than when using raidz or raidz2. When the system gets to the point of being full enough to risk increased fragmentation, slower writes, or running out of space, it can then be converted to use normal dual mirrors. The redundant mirrors are then remapped to become new mirror vdevs. The main drawbacks to this approach is that the maximum total storage space is less than when using raidz or raidz2, and the data already written at the time of transition is not as optimally written as if dual mirrors were used from the start. Has anyone used this approach in practice? Bob ====================================== Bob Friesenhahn [EMAIL PROTECTED], http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/ _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
