Question & Answer
What are the differences between using raw disk vs. cooked files for informix chunk files on a UNIX server?
Advantages of raw disk partitions:
- faster I/O without UNIX file system overhead,
- more synchronous I/O without UNIX file system buffering,
- more reliable without potential for UNIX file system corruption.
- may require re-partition of disk to match calculated chunk space sizes,
- root (super-user) access required to repartition disk.
- If using cooked files, Informix strongly recommends not using NFS-mounted file systems, and using local physical disks only.
- When specifying the raw disk partition device names to use, use the character-special (raw, unbuffered) device named something like /dev/rxxx, rather than the block (cooked, buffered) device named /dev/xxx.
- Using offsets vs. multiple raw disk partitions: When using raw disk partitions, more than one chunk space can be put on the same raw disk partition if the subsequent chunk spaces are offset from the previous ones. It's possible to place all three standard chunk spaces on a single raw disk partition. The disadvantage of this is that you may forget where you configured your database server's chunk spaces.
If you used cooked files, or one raw disk partition per chunk space, you can always enter " ccmsrv status -d " (Synergy 5.x/6.x) to report where the chunk spaces are located. If you configured multiple chunk spaces on one raw disk partition, these commands do not report what the offsets are.
Using offsets: When specifying the number of kbyte blocks to offset, this offset must be a multiple of the system's page size. The system page size varies depending on platform. On a system with a 2048 byte (2K) page size, you would not be allowed to specify an odd numbered offset.
16 June 2018