19.1.1 File sizes required for the execution of the database load utility (pdload)

The following table shows the formulas for determining the file sizes required for the execution of the database load utility (pdload):

File typeFormula (bytes)
Input data fileh x b
Index information file
B-tree index:
(d + y) x (b + e) + 512
Plug-in index:
(12 + q) x p + 1,024
These formulas are for the size of one index. If there are multiple indexes, determine the size of each index.
Error information filek x f + s x 200
Temporary file for creating error information fileIn the following conditions, the work file output directory will need number-of-key-duplication-errors x 8 + number-of-errors-detected-by-plug-in-function x 200 of space for each server that has a table storage RDAREA. For details about work file output destination directories, see 2.3.3 Creating a work file output directory.
  • For a HiRDB/Parallel Server, the server that contains the input files is different from the server containing the table storage RDAREAs.
LOB input fileb
[Figure] (LOB data length + 4)i
i=1
LOB middle fileB
[Figure]{
i=1
c
[Figure] (LOB file name length-ij + 36) + 24} + 1,024 + c x 84
j = 1
Error data fileMIN(f, g) x h
Process results file1,500 + number of servers storing table x 500
Work file#[4 + 2 x R + 2 x r + 4 x I x R + {b[Figure] (value of past message output interval specified by the -m option)}] x 200
Work file for sorting
Condition 1:
Size of index information file + 4 x (b + e)
Condition 2:
{Size of index information file + 4 x (b + e)} x 2
  • Condition 1
    When the work buffer size specified in the sort statement [Figure] E
  • Condition 2
    When the work buffer size specified in the sort statement < E
E: Buffer size
The buffer size obtained according to buffer-size-for-sorting in Database Load Utility (pdload) of the manual HiRDB Version 9 Command Reference.
a: Number of input rows x number of LOB columns
b: Number of input rows (for a repetition column, number of input rows x number of elements)
c: Number of LOB columns
d: Index key length
See Table 16-5 List of index key lengths. For variable-length data, treat a single column as multicolumn and use the largest defined length.
e: Number of existing rows (for a repetition column, number of existing rows x number of elements)
f: Number of error data items
g: Number of output rows specified in the errdata operand of the source statement
h: Average source record length
k: If there is a column with an abstract data type, the value is 300; otherwise, it is 120.
m: For a DAT-format file or a binary format file output by pdrorg, the value is 0.
For any other file, the value is (record length of one row in the input file x 4).
p: If index storage RDAREAs are initialized, the value is (b + e); otherwise, the value is b.
q: Value as follows
  • 27 for the abstract data type stored in the LOB RDAREA
  • Key length + 2 for the abstract data type of a maximum of 255 bytes of definition length
  • 2 for the abstract data type of 256 bytes or more of definition length
Typical abstract data type values are as follows.
  • 27 for the SGMLTEXT type
  • 2 for the FREEWORD, GEOMETRY, and XML types
r: Number of RDAREAs for LOB storage
s: Number of servers
y: If all key component columns are fixed length, the value is 10; if they include a variable length, the value is 12.
I: Number of indexes
R: Number of partitioned index or table RDAREAs
Note
When calculating the size of index information files and sort work files, if the index configuration columns are repetition columns, b and e do not refer to the number of rows but to (number of rows x number of elements).
#
Output if lvl2 is specified as the information message output suppression level in the -m option.