This section lists some general strategies for disk layout that do not require any information about the characteristics of a particular database.
When you allocate disk space (buffered or unbuffered files), you allocate it in chunks. A dbspaceis associated with one or more chunks. You must allocate at least one chunk for the root dbspace.
It is recommended that you format your disks so that each chunk is associated with its own disk partition. You can easily track disk-space use when you define every chunk as a separate partition (or device). You can also avoid errors caused by miscalculated offsets.
A disk that is already partitioned might require the use of offsets. For details, see Allocating Raw Disk Space on UNIX.
Mirror the critical dbspaces: the root dbspace, the dbspace that contains the physical log, and the dbspace that contains the logical-log files. You specify mirroring on a chunk-by-chunk basis. Locate the primary and the mirror chunk on different disks. Ideally, different controllers handle the different disks. Figure 63 shows a primary chunk and its mirror.
You can use the DBSPACETEMP environment variable and configuration parameter to store a list of dbspaces used for temporary storage. The list can include both temporary and standard dbspaces. To achieve load balancing, design the list so that your temporary disk space is spread across multiple disks. For instructions on how to set DBSPACETEMP, see the chapter on configuration parameters in the IBM Informix: Administrator's Reference.
Whether or not you use logging tables, the logical log and physical log both contain data that the database server accesses frequently. Reserved pages are also accessed frequently; they contain internal tables that describe and track all dbspaces, chunks, databases, and tblspaces.
When you allocate disk space (buffered or unbuffered files), you allocate it in chunks. A dbspace is associated with one or more chunks. You must allocate at least one chunk for the root dbspace.
By default, the database server stores the logical and physical logs together with the reserved pages in the root dbspace. Storing the logical and physical logs together is convenient if you have a small, low-volume transaction-processing system. However, maintaining these files together in the root dbspace can become a source of contention as your database system grows.
To reduce this contention and provide better load balancing, move the logical and physical logs to separate partitions or, even better, separate disk drives. For optimum performance, consider creating two additional dbspaces: one for the physical log and one for the logical log. When you move the logs, avoid storing them in a dbspace that contains high-use tables. Instead, consider storing them in a dbspace dedicated to storing only the physical or logical log. For more advice on where to store your logs, see Location of the Physical Log and Location of Logical-Log Files.
For instructions on how to change the location of the logical and physical log, see Using a Text Editor to Change Physical-Log Location or Size and Moving a Logical-Log File to Another Dbspace.
When you plan your disk layout, consider how the configuration that you choose affects your backup-and-restore procedure. This section describes two configurations that can have a significant impact on your backup-and-restore procedure.
When a disk that contains the system catalog of a particular database fails, the entire database remains inaccessible until you restore the system catalog. It is recommended that you do not cluster the system catalog tables for all databases in a single dbspace but instead place the catalogs with the data that they track.
Although it makes sense from a performance perspective to separate the root dbspace from the physical and logical logs, and the two logs from one another, this configuration is the least desirable in terms of recovery.
Whenever a disk that contains critical information (the root dbspace, physical log, and logical log) fails, the database server comes offline. In addition, the database server administrator must restore all the database server data, starting in offline mode, from a level-0 backup before processing can continue.
When you separate the root dbspace from the physical- and logical-log files, you increase the probability that, if a disk fails, it is one that contains critical information (either the root dbspace, physical log, or logical log). For information on how to fragment to improve backup and restore characteristics, see your IBM Informix: Performance Guide.
Home | [ Top of Page | Previous Page | Next Page | Contents | Index ]