9.2.1 Designing HiRDB file system areas for RDAREAs

This section discusses the design considerations for HiRDB file system areas in which RDAREAs are to be created.

Organization of this subsection
(1) Design for improved reliability
(2) Design for improved performance

(1) Design for improved reliability

  1. For update processing, character special files are more reliable than regular files. Regular files may become inaccessible if the OS terminates abnormally. Thus, character special files should be used for HiRDB file system areas for user RDAREAs that satisfy the following characteristics:
    • User RDAREAs for storing tables that are used primarily for update processing.
    • User RDAREAs for storing particularly important data.
  2. The amount of space required for a HiRDB file system area is displayed when the system generator (pdgen) is executed. At least the displayed amount of space should be allocated to the HiRDB file system area.
  3. HiRDB file system areas for RDAREAs must be created at the server machine where the single server is defined.
  4. For the following RDAREAs, create HiRDB file system areas on the server machine where the single server is defined:
    • System RDAREAs
    • Data dictionary LOB RDAREAs
    • Registry RDAREAs
    • Registry LOB RDAREAs
  5. For the following RDAREAs, create HiRDB file system areas on the server machine where the single server is defined:
    • User RDAREAs
    • User LOB RDAREAs
  6. If the system switchover facility is to be used, the HiRDB file system areas for user RDAREAs should be allocated as character special files.

(2) Design for improved performance

  1. You should create separate HiRDB file system areas for the following types of RDAREAs:
    • System RDAREAs
    • Data dictionary LOB RDAREAs
    • User RDAREAs
    • User LOB RDAREAs
    • Registry RDAREAs
    • Registry LOB RDAREAs
  2. You should create HiRDB file system areas for system files on hard disks separate from the ones used for HiRDB file system areas for RDAREAs. In this way, you can distribute input/output operations when collecting a synchronization point dump, thereby reducing the amount of time required to collect the synchronization point dump.
  3. If you are not using the prefetch facility, the sequential read speed is faster with regular files than with character special files.
  4. For random one-page reads, processing speed is faster with character special files than with regular files.
  5. For write processing, the speed is faster with character special files than with regular files.
  6. Regular files have a hierarchical structure; therefore, as files become larger, the hierarchical levels increase. When you access such files with many hierarchical levels, the number of input/output operations increases, adversely affecting the access efficiency.
  7. The HiRDB file system areas should be allocated as shown in Table 9-1, so that input/output time can be reduced.

    Table 9-1 Allocation of HiRDB file system areas to improve performance

    Type of HiRDB file system areasFile to be allocated
    HiRDB file system area for system RDAREAsCharacter special file
    HiRDB file system area for dictionary LOB RDAREAs
    HiRDB file system area for user LOB RDAREAs
    HiRDB file system area for user RDAREAs that store frequently updated tables or tables primarily subject to retrieval of small amounts of data
    HiRDB file system area for user RDAREAs that store tables with a large amount of data that are primarily subject to retrieval of all entries or retrieval of large amounts of data by key using the cluster key (but with data that is rarely updated)
    • Regular file (when not using the prefetch facility)
    • Character special file (when using the prefetch facility)