I'm not such an FS expert, so please step in to correct me in case I am wrong.
All large databases (Oracle, UDB, MSSQL) have two different access patterns for files. One type is data access, one type is log file access. The question is now obviously divided into two questions: Which FS should you use for each type of files? Log files are ALWAYS written to sequentially, and very, very, VERY RARELY read from (usually only when recovering from crashes). I'd go with a journalling FS here. Selecting the most efficient one is left as an exercise for the reader. Data file access is generally random-access while reading (when you need a data block, you need it *now*, usually), and "elevator" access while writing (since the DB chooses when to write data blocks to disk, and does so in a LBA-friendly manner). There are some optimizations which break both those assertions (multi-block read IO and block write pinging in cluster machines), but let's leave those out of the picture for now. Here I would go with raw block devices. if you are unable to go with raw block devices (e.g. because you cannot predict data size, you need data portability (i.e. Oracle's transportable tablespaces)), go with a non-journalling FS, but only since no journalling FS does compaction of files (yet). If they had big-blocked (e.g. 1 block = 16MB) FAT16, with multiple "super-blocks", I would suggest that one. If a journalling FS had some user-space tool for that, and you had a setup that actually allows you to perform a "defrag" (a 24x7 setup probably won't allow it), they you should consider it. However, as these are not available AFAIK, the discussion is rather moot. Just my $0.02 ... Shachar Tal Verint Systems -----Original Message----- From: Maxim K. [mailto:[EMAIL PROTECTED] Sent: Sunday, October 19, 2003 4:50 PM To: Linux-IL Mailing List Subject: filesystem for database box Hello, Linux People! after short consultation, i have come to this conclusion about which filesystem i should use on my database box (or server) the winner is: ex2 linux extended filesystem, yes, lassies & lads - "why not ext3/Reiser's ?" - "because journalling is already implemented in the DBMS." - "why not jfs then ? - "it is too young, my friends + journalling..." - "why not xfs , it is so ... juicy ? - "it is too resource demanding (i have p2 670MHz + 256 MB )" if somebody thinks otherwise, please tell me why! thank you. ================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED] This electronic message contains information from Verint Systems, which may be privileged and confidential. The information is intended to be for the use of the individual(s) or entity named above. If you are not the intended recipient, be aware that any disclosure, copying, distribution or use of the contents of this information is prohibited. If you have received this electronic message in error, please notify us by replying to this email. ================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED]
