> I've read about using symbolic links, but only for using > multiple data > directories, not splitting a file (not to mention, across > more than one node).
Symlinks wont work for splitting a file... > Since MySQL doesn't directly support this, any ideas on doing > it at the OS > level to 'trick' MySQL into believing its all in a single > directory as a single Ironically enough, I was at one point looking for a Linux kernel module that would have done something fairly close to this... Never got it anywhere NEAR done though. However, even if you *could* split the data file at the OS level, whatever did the splitting would need to be aware of how the data is structured and split it in a meaningful way. Simply chopping the file into N pieces would be worse than useless (you'd very likely wind up with a row that was partly in one file and partly in another, plus you wouldn't be able to reliably predict which rows were in which partitions). > table? It's looking more like I may have to build a DBsmart > API to select the > partition/machine containing the data I need. I recall a > group of sharpsters > at Temple U. did something similar to that. I'm unfamiliar with any such project. However you still have the problem of naming collisions. If you want the table to be named "foo" regardless of which partition it's in, and the master has copies of all partitions -- how would you arrange things on the master to avoid the naming collision? Thus we come back to using distinct table names for the partitions. On the master you can have a table "foo" which is the logical aggregate of all the partitions, and on any given machine you can have a table "foo" which is the logical aggregate of all the partitions that exist on that particular machine if you use the RAID table type. Alternatively you could consider a database like DB2 which supports seamlessly partitioning a table across N machines in a DB cluster based on the value of a key column (the "portioning key"). In such an arrangement, all rows with value X exist on server A, all rows with value Y exist on server B, and any process accessing any server will see one table that contains all rows (accessing rows from another server will be slower of course). -JF -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]