While researching to buy/build an HTPC (likely to run XBMC) I ran across: http://forum.xbmc.org/showthread.php?s=25935987f91ef317c33806af17e06e7a&t=94268
unRAID Advantages of Unraid -Allows you to mix and match drives of different sizes and makes into a single "User Share" of pooled storage (one big folder across drives) -Allows you to pull the drives of the array out and read the data on them on another computer -Allows drives to spin down if they are not being accessed, saving power and possible prolonging their life -Unraid allows for the growing of the array in size by replacing one drive at a time with full use of that drive after addition and no data loss -Unraid is easy to install, runs off a pen drive to save sata ports for data, and is configured by an easy to use web interface Downsides: -Unraid costs money for real versions -Unraid's write speeds are pretty low without a cache drive -Unraid's read speeds are slightly lower than the drives by themselves, but still more than fast enough for media -Unraid is a dedicated NAS OS, meaning that it can't easily do other server stuff for you In the end Unraid gives you a media server that is grown periodically as storage is needed, one cheapest disk available at a time, yet it can easily stream multiple HD movies to multiple (I have tested up to 5) clients on your network. Here is a recommended Unraid build: http://forum.xbmc.org/showthread.php...ghlight=unraid And from the vendor's site: http://www.lime-technology.com/technology Our approach uses a storage organization we call unRAID(TM). unRAID(TM) is similar to RAID-4 in that for every n hard drives, there are n-1 data drives, and a single fixed parity drive. This provides similar drive failure protection as other RAID organizations: * Tolerates a single failed hard drive. Array parity permits reconstruction of a single failed data disk. * Tolerates single drive read errors . Read errors from a single disk are corrected "on-the-fly". Unlike other RAID organizations, however, files are not striped across the data drives. Instead, each data drive is formatted normally with its own file system. The unRAID(TM) organization offers several key advantages over other RAID systems: * Increased flexibility: not all hard drives need to be the same size or speed. The only requirement is that the parity drive must be as large as or larger than any of the data drives. * Easier expansion: new data drives may be added to the array without disturbing the data on the other drives. Since data is not striped, no lengthy stripe reorganization needs to take place before the new storage is available. * Better fault tolerance: in the unlikely event of catastrophic hardware or software failure, data corruption would be much more isolated than that which could occur with other RAID organizations. It is almost impossible to lose all your data, since each individual data drive has its own file system. * Better power management: not all hard drives are required to be spinning in order to access data normally; hard drives not in use may be spun down. * Potentially better multiple-stream read performance. If multiple files are read from different disks, overall read performance can be higher than with other RAID organizations. Sound like it has a lot of similarities to Data Robotics' Drobo (http://www.drobo.com/), which similarly lets you add drives ad-hoc. Ever since the Drobo came out (~2007) I've been expecting to see the equivalent functionality turn up as open source. That the XBMC people are recommending unRAID - another closed source commercial solution - implies that this hasn't happened yet. (ZFS pools are perhaps the closest approximation: http://en.wikipedia.org/wiki/ZFS#Storage_pools but I don't think you can pull an arbitrary drive from a pool and replace it with a higher capacity one and have the storage system incorporate the new drive without any further action. Though perhaps all that is needed is a bit of automation glue.) A traditional NAS (or SAN) is fine for a corporation with a dedicated storage admin who can plan out expected growth, afford to overbuy capacity, and take the time to backup and rebuild a RAID with higher capacity drives when needed. If performance is critical, it makes sense to tune your RAID array. The ability to arbitrarily swap out drives for higher capacity drives is a "killer feature" for a NAS. No longer is the NAS a static chunk of storage, but instead something that can be incrementally expanded with almost no effort. You are only limited by the number of drive slots you have and the size of the disks available. And anyone capable of mechanically swapping a drive is able to perform the upgrade (just make sure they don't pull out more than one drive). Drobo seemed to get a lukewarm reception, and I don't hear much about it these days. Part of that was because their initial product used a USB interface rather than being a NAS. Later they offered an add-on to make it a NAS which was not competitively priced. And it was consistently panned for its performance. It seems unRAID has worked through some of these issues. Minimizing the performance hit (you give of striping, but otherwise maintain single drive performance), and offering a software-only option so you can minimize hardware costs. Looks like the pricing for unRAID is: Free - 3 drives, no cache drive $70 - 6 drives + cache drive $120 - 21 drives + cache drive Oddly their marketing makes no mention of the protocols used to share files, but does make a few references to "My Network Places," implying SMB/CIFS. (And here I thought we were moving away from a Windows-centric world.) The FAQ on their Wiki notes: http://lime-technology.com/wiki/index.php?title=FAQ#How_do_I_configure_NFS_mounts.3F Starting with unRAID 4.4.1, unRAID includes the ability to serve files using the NFS protocol. This release includes NFS export ability for disk and user shares. ... Please consider NFS support to be experimental. For example, there has been no performance tuning done whatsoever (as of 4.5beta1/4.4.2). However, see this thread for improved settings and better performance. This is a good illustration of why we'd be better off getting unRAID-like functionality added to a solid, widely used OS, like Linux or FreeBSD. Though if you dig through the FAQ it does say unRAID is "based on Slackware Linux" w/kernel 2.6. Anyone tried out unRAID? Anyone tried building the equivalent with ZFS or some other open source technology? -Tom -- Tom Metro Venture Logic, Newton, MA, USA "Enterprise solutions through open source." Professional Profile: http://tmetro.venturelogic.com/ _______________________________________________ Discuss mailing list Discuss@blu.org http://blu.org/mailman/listinfo/discuss