Nonstop Database, HiRDB Version 9 Installation and Design Guide

[Contents][Index][Back][Next]

19.2.3 Memory size required for the execution of the database load utility (pdload)

The following tables show the formulas for determining the memory size required for the execution of the database load utility (pdload). For details about the variables, see (3) Variables used in the formulas.

Organization of this subsection
(1) HiRDB single server configuration
(2) HiRDB parallel server configuration
(3) Variables used in the formulas

(1) HiRDB single server configuration

Condition Formula for determining memory size (KB)
32-bit mode Single server 6,352 + [Figure]{([Figure] + [Figure]) [Figure] 1,024}[Figure]
Utility special unit# 1,968 + [Figure]{([Figure] + [Figure]) [Figure] 1,024}[Figure]
64-bit mode Single server 12,172 + [Figure]{([Figure] + [Figure]) [Figure] 1,024}[Figure]
Utility special unit# 2,423 + [Figure]{([Figure] + [Figure]) [Figure] 1,024}[Figure]

#: If no utility special unit is being used, use the value for a single server.

(2) HiRDB parallel server configuration

Condition Formula for determining memory size (KB)
32-bit mode MGR 2,255 + [Figure]{[Figure] [Figure] 1,024}[Figure]
Server machine containing input files 2,045 + [Figure]{([Figure] + [Figure]) [Figure] 1,024}[Figure]
BES# 3,762 + [Figure]{[Figure] [Figure] 1,024}[Figure]
64-bit mode MGR 2,575 + [Figure]{[Figure] [Figure] 1,024}[Figure]
Server machine containing input files 2,406 + [Figure]{([Figure] + [Figure]) [Figure] 1,024}[Figure]
BES# 8,247 + [Figure]{[Figure] [Figure] 1,024}[Figure]

#: If a single server machine has multiple back-end servers, add into the calculation only for the number of back-end servers.

(3) Variables used in the formulas

[Figure] (bytes):
{3,056 + A + B + (516 [Figure] a) + (572 [Figure] b) + (312 [Figure] c) + (144 [Figure] d) + (8 [Figure] e) + (1,032 [Figure] f) + (44 [Figure] g) + (272 [Figure] h) + (224 [Figure] i) + (44 [Figure] j) + (60 [Figure] k) + (260 [Figure] m) + (56 [Figure] n) + (196 [Figure] p) + (236 [Figure] q) + (744 [Figure] r) + (620 [Figure] s)} [Figure] 2

[Figure] (bytes):
{6,908 + [Figure] + (C [Figure] t) + K + (48 [Figure] a) + (22 [Figure] b) + (8 [Figure] e) + (240 [Figure] i) + (48 [Figure] j) + (4 [Figure] k) + (224 [Figure] m) + (47,416 [Figure] t) + (1,032 [Figure] u) + (4 [Figure] v)}

[Figure] (bytes):
{37,700 + ([Figure] [Figure] 2) + C + D + F + H + P + Q + T + (80 [Figure] a) + (1,871 [Figure] b) + (120 [Figure] c) + (26 [Figure] g) + (1,532 [Figure] i) + (36 [Figure] j) + (44 [Figure] k) + (1,212 [Figure] m) + (40 [Figure] n) + (344 [Figure] p) + (30 [Figure] q) + (16 [Figure] u) + (88 [Figure] v) + (20 [Figure] w)}

[Figure] (bytes):
{69,436 + [Figure] + D + K + E + L + M + N + S + (U [Figure] 2) + 8 + (48 [Figure] a) + (32 [Figure] a) + (88 [Figure] c) + (4 [Figure] g) + (2,156 [Figure] k) + (24 [Figure] t) + (1,024 [Figure] u) + (4 [Figure] v) + (50 [Figure] y) + (50 [Figure] z)}

a: Number of columns

b: Number of columns with abstract data type

c: Number of parameters in constructor or reverse constructor function

d: Number of file path names specified in command line or control information file

e: Number of LOB middle files

f: Number of LOB files by the column that are specified

g: Number of table storage RDAREAs

h: Number of table row partitioning conditions

i: Number of indexes

j: Number of index storage RDAREAs

k: Number of BLOB-type columns

m: Number of plug-in indexes

n: Number of user LOB RDAREAs storing LOB-attribute abstract data type

p: Number of functions provided by plug-in

q: Number of function parameters provided by plug-in

r: Number of data-type plug-ins

s: Number of index-type plug-ins

t: Number of servers storing tables

u: Number of BLOB-type parameters among the constructor function parameters used

v: Number of user LOB RDAREAs storing LOB columns

w: Number of user LOB RDAREAs storing plug-in indexes

y: Number of BINARY columns
This is the BINARY-data columns in input data that have been excluded from processing by the skipdata control statement plus the number of columns actually defined in the tables.

z: Number of BINARY-attribute parameters for plug-in-provided functions
This is the number of BINARY data columns of the plug-in-provided functions in input data that have been excluded from processing by the skipdata control statement, plus the number of columns actually defined in the tables.

A: Total file size specified in the command line

B: Total length of file path names specified in the command line and control information file

C: If the following condition is satisfied, the value is (pd_utl_buff_size [Figure] 1,024 + 4,096) [Figure] 2; otherwise, the value is 0:
  • HiRDB single server configuration
    A utility special unit is used.
  • HiRDB parallel server configuration
    The pdload command is executed with a back-end server name containing the table storage RDAREA that is different from the server name specified in the source statement; or, the pdrorg command specifying the -g option is executed.

D: Row length
This is the sum of the defined lengths of all columns that constitute the table. The length of a BINARY-type row is the defined length for pdload plus MIN(defined length, 32,500) for pdrorg (in bytes). For a non-FIX table, add (a + 1) [Figure] 4.

E: Memory required for EasyMT
Add this value if you have specified easymt in the -f option.

F: 550 [Figure] 1,024 + 1,024 [Figure] 1,024 + G
Add this value if you have specified c in the -i option.

G: 256 (32-bit mode) or 512 (64-bit mode)

H: Value specified for the batch input/output local buffers [Figure] RDAREA page length [Figure] J + value specified for the random access local buffers [Figure] RDAREA page length
Add this value if you have specified the -n option. If an RDAREA's page length varies from one partitioned RDAREA to another, use the longest page length for this calculation.

J: Determine from the following table:
Value specified with -n option Table partition type
Non- partitioning table Key range partitioning table Hash partitioning table
Rebalancing hash
(HASHA-HASHF)		 Non- rebalancing hash
(HASH0-HASH6, HASHZ)
FIX hash Flexible hash
div specified 1 Number of row partitions in the server for the table HiRDB single server configuration: 1,024
HiRDB parallel server configuration: ([Figure]1,024 [Figure] g[Figure]) [Figure] (number of table storage RDAREAs on the server)		 Number of row partitions in the server for the table Number of row partitions in the server for the table
div not specified 1 1

K: Parameter length
Sum of the lengths of arguments in the constructor function that is used to generate values for the abstract data type. For a BLOB-type parameter, the parameter length is 8 bytes.

L: If all the following conditions are satisfied, add 1 or 2; otherwise, the value is 0:
  • The errdata operand is specified in the source statement.
  • A utility special unit is used (HiRDB single server configuration), or the server name specified in the source statement is not the name of the back-end server that contains the table storage RDAREA (HiRDB parallel server configuration).
  • The table contains an abstract data type column, or a unique index is defined.
  1. errwork operand specified: Value of errwork operand [Figure] 1,024
  2. errwork operand omitted: Value of pd_utl_buff_size [Figure] 1,024 [Figure] 3 [Figure] t

M: Memory required for UOC
Add this value if you use UOC.

N: When the maxreclen operand is specified, calculate the following value:
If the input data file is in extended DAT format:
Value specified in maxreclen operand [Figure] 1,024
If the input data file is in DAT format:
Value specified in maxreclen operand [Figure] 1,024 [Figure] 3
When the table has BINARY type columns, or when the input data files are in binary format:
Add the smaller of the following to the calculation:
  • Value specified in maxreclen operand [Figure] 1,024
  • Variable D (row length)
When other than the above:
0

P: Memory required for plug-ins
Add this value if there is an abstract data type column provided by a plug-in. For details about the size of memory required by plug-ins, see the applicable plug-in documentation.
If the constructor functions' arguments are BLOB or BINARY type, add (actual parameter length stored in all abstract data types [Figure] 2) that is defined per row.

Q: Memory requirement for output buffer
If the specified index creation method is the batch index creation mode or the index information output mode and the following condition is satisfied, add 2 megabytes:
  • Number of table partitions [Figure] number of index definitions > maximum number of processes that can be open - 576

S: (4 [Figure] Buffer length + 1.1) [Figure] 1,024
The buffer length is as follows (round up the value in units of 32 kilobytes).
  • The value specified in the option statement file_buff_size of the database load utility (pdload).
  • If the above value is not specified: 1,024.

T: If there are compressed columns, split compression size [Figure] 2 + RDAREA's page length; if there are no compressed columns, 0
For the split compression size, use the largest value among all the compressed columns. If the page length varies from one RDAREA to another for a row-partitioned table, use the longest page length for this calculation.

U: Row length of input file (bytes)
DAT format: Value specified for the maxreclen operand [Figure] 1,024 (or 32,768 if the maxreclen operand is not specified)
Fixed-length format: Row length of input file
Binary format: 0

[Contents][Back][Next]

[Trademarks]

All Rights Reserved. Copyright (C) 2012, 2015, Hitachi, Ltd.