obnoxio
User
Baby Cheetah
| Posts: 8 |   |
Karma: 1
|
Re:I am new... - 2007/02/22 20:29
It's easy enough, though, just type in the word.
Lazy git! 
Informix uses pools of disk for storage, in general it performs better on raw disk slices. In order to let the database (as a server) know where it can store on disk, IDS defines the concept of a dbspace. The dbspace consists of one or more chunks, where a chunk is an arbitrary allocation of disk space, which can be either a raw disk slice, a "cooked" disk slice, or a flat file in the filesystem (which gives sucky performance). If a dbspace starts getting full, you can add more chunks to it. Chunks can mix between raw, cooked and file, but performance will vary accordingly.
Your tables are then stored somewhere in the dbspaces (you can control which dbspace[s] are used for a table, but not which chunk within the dbspace.) The server pre-allocates a small piece of disk within the dbspace to the table, called an initial extent. When that extent gets full, a "next extent" will be allocated. If you have lots of non-contiguous extents, disk IO can suffer as the head moves around all over the disk. In an ideal world, you will make the extents big enough to load all the data in the table into no more than one or two dbspaces, but more is not inherently bad. It is bad for performance when you have hundreds of little extents interleaved with thousands of extents from other tables.
|
