Nonstop Database, HiRDB Version 9 Installation and Design Guide

19.2.13 Memory size required for the execution of the rebalancing utility (pdrbal)

(1) For a HiRDB/Single Server

Formula for determining memory size (KB)

8,756#3 + 536 + 0.02 x number of columns + 0.2 x number of target RDAREAs + 1.7 x number of source RDAREAs + 0.26 x number of target RDAREAs x number of indexes + (0.09 + average index statement file length 1,024) x number of index statements + (0.02 + average directory length 1,024) x (number of idxwork statements + number of sort statements) + (length of control information file + length of execution results file) 1,024 + 0.05 x number of columns + 0.05 x number of RDAREAs + 0.15 x number of indexes + 0.05 x number of index storage RDAREAs + 550 x 1,024 + size of work file for sorting#1 -n option specified: + page length of RDAREA#2 x number of batch input/output buffer sectors x y Applicable table containing LOB columns: + 64 + 0.01 x number of LOB columns + 0.18 x number of target RDAREAs + 0.09 x number of source RDAREAs + 0.08 x number of LOB storage RDAREAs When the target table contains BINARY columns: + 33 x (number of BINARY columns x number of target RDAREAs x number of source RDAREAs) + buffer length used for BINARY columns#5 Applicable table containing an abstract data type provided by plug-in: + 40 + (0.27 + 2 x length of abstract data type) x number of abstract data-type columns + 0.3 x number of unld_func statements + (128 + 0.11 x number of LOB attributes + 0.1 x number of functions specifying unld_func + 0.07 x number of abstract data type attributes) x number of abstract data-type columns + (33 x number of BINARY attributes x number of target RDAREAs x number of source RDAREAs) x 2 + 0.01 x number of plug-in indexes + 0.19 x number of unld_func statements + (average length of unld_func statements 1,024 x number of unld_func statements) + (average length of reld_func statements 1,024 x number of reld_func statements) + number of abstract data-type columns x 1 + (number of LOB attributes x 0.05) x number of RDAREAs + number of data type plug-ins x 10 + number of plug-in indexes x 10 + memory required for plug-ins When the specified index creation method is the batch index creation mode or the index information output mode and the following condition is satisfied: Number of table partitions x number of index definitions > maximum number of processes that can be open - 576 + 2,048 When there are compressed columns: + split compression size#4 x z + RDAREA's page length#2

y: Use one of the following values:

• When the rebalancing facility is used with a FIX hash-partitioned table
  (1,024 number of storage RDAREAs for the entire table) x number of table storage RDAREAs in the corresponding server
  
• Other than the above
  1
  

z: Use one of the following values, as applicable:

• Exclusive mode (-k exclusive): 2
• Shared mode (-k share): 1

#1: Add this value during batch index creation (-ic specified or omitted).

#2: If the page length varies from one RDAREA to another for a row-partitioned table, use the longest page length for this calculation.

#3: For the 64-bit mode, the value is 14,272.

#4: Use the largest value among all compressed columns for the split compression size.

#5: Use one of the following values, as applicable, for the buffer length used for BINARY columns:

• Exclusive mode
  0
  
• Shared mode
  _n
  (definition length i)
  ^{i = 1}
  n: Number of BINARY columns
  

(2) For a HiRDB/Parallel Server

Condition	Formula for determining memory size (KB)
MGR	1,498#3 + 2 + 0.05 x number of columns + 0.05 x number of RDAREAs + 0.15 x number of indexes + 0.05 x number of index storage RDAREAs + (0.09 + average index statement file length 1,024) x number of index statements) + (0.02 + average directory length 1,024) x (number of idxwork statements + number of sort statements) + (length of control information file + length of execution results file) 1,024 Applicable table containing LOB columns: + 0.08 x number of LOB storage RDAREAs Applicable table containing an abstract data type provided by plug-in: + 0.19 x number of unld_func statements + (average length of unld_func statement 1,024 x number of unld_func statements) + (average length of reld_func statement 1,024 x number of reld_func statements)
DS	1,455#4 + 32 + 0.33 x number of target BESs + 0.3 x number of source BESs + 0.2 x number of target RDAREAs + 0.22 x number of source RDAREAs + 0.34 x number of FESs + (0.09 + average index statement file length 1,024) x number of index statements + (0.02 + average directory length 1,024) x (number of idxwork statements + number of sort statements) + 0.05 x number of columns + 0.05 x number of RDAREAs + 0.15 x number of indexes + 0.05 x number of index storage RDAREAs Applicable table containing LOB columns: + 0.08 x number of LOB storage RDAREAs When the target table contain BINARY columns: + 33 x (number of BINARY columns x number of target RDAREAs x number of source RDAREAs) Applicable table containing an abstract data type provided by plug-in: + 0.01 x number of target BESs + 0.19 x number of unld_func statements + (average length of unld_func statements 1,024 x number of unld_func statements) + (average length of reld_func statements 1,024 x number of reld_func statements) + number of abstract data type columns x 1 + (number of LOB attributes x 0.05) x number of RDAREAs + number of data type plug-ins x 10 + number of plug-in indexes x 10
BES	6,601#5 + 50 + (517 + 0.01 x number of columns) x number of target BESs + (33 + 0.01 x number of columns) x number of source BESs + 0.2 x number of target RDAREAs + 1.7 x number of source RDAREAs + 0.01 x number of columns + 0.26 x number of target RDAREAs x number of indexes + (0.09 + average index statement file length 1,024) x number of index statements + (0.02 + average directory length 1,024) x (number of idxwork statements + number of sort statements) + 0.05 x number of columns + 0.05 x number of RDAREAs + 0.15 x number of indexes + 0.05 x number of index storage RDAREAs + 550 x 1,024 + size of work file for sorting#1 -n option specified: + page length of RDAREA#2 x number of batch input/output buffer sectors x y Applicable table containing LOB columns: + 32 + 0.01 x number of LOB columns + (32 + 0.01 x number of LOB columns) x number of target BESs + 0.18 x number of target RDAREAs + 0.1 x number of source RDAREAs + 0.08 x number of LOB storage RDAREAs When the target table contain BINARY columns: + 33 x (number of BINARY columns x number of target RDAREAs x number of source RDAREAs) + buffer length used for BINARY columns#7 When the target table contains abstract data types provided by plug-ins: + 40 + (0.27 + 2 x length of abstract data type) x number of abstract data-type columns + 0.3 x number of unld_func statements + {(64 + 0.05 x number of LOB attributes) x number of abstract data-type columns} x number or target BESs + {(64 + 0.01 x number of functions specifying unld_func + 0.07 x number of abstract data type attributes + 0.05 x number of LOB attributes) x number of abstract data-type columns} x number of source BESs + (33 x number of BINARY attributes x number of target RDAREAs x number of source RDAREAs) x 2 + 0.01 x number of plug-in indexes + 0.19 x number of unld_func statements + (average length of unld_func statements 1,024 x number of unld_func statements) + (average length of reld_func statements 1,024 x number of reld_func statements) + number of abstract data-type columns x 1 + (number of LOB attributes x 0.05) x number of RDAREAs + number of data type plug-ins x 10 + number of plug-in indexes x 10 + memory required for plug-ins When the specified index creation method is the batch index creation mode or the index information output mode and the following condition is satisfied: Number of table partitions x number of index definitions > maximum number of processes that can be open - 576 + 2,048 When there are compressed columns: + split compression size#6 x z + RDAREA's page length#2

y: Use one of the following values:

• When the rebalancing facility is used for FIX hash partitioning tables
  (1,024 number of storage RDAREAs for the entire table) x number of table storage RDAREAs in the corresponding server
  
• Other than the above

z: Use one of the following values, as applicable:

• Exclusive mode (-k exclusive): 2
• Shared mode (-k share): 1

#1: Add this value during batch index creation (-ic specified or omitted).

#2: If the page length varies from one RDAREA to another for a row-partitioned table, use the longest page length for this calculation.

#3: For the 64-bit mode, the value is 4,660.

#4: For the 64-bit mode, the value is 15,830.

#5: For the 64-bit mode, the value is 11,804.

#6: Use the largest value among all compressed columns for the split compression size.

#7: Use one of the following values, as applicable, for the buffer length used for BINARY columns:

• Exclusive mode
  _m
  ((definition length i)
  ^{i = 1}
  _n
  + (definition length i x 9)) x number of target BESs
  ^{i = 1}
  m: Number of BINARY columns for which the compression specification is not specified
  n: Number of BINARY columns for which the compression specification is specified
  
• Shared mode
  _n
  (definition length i) x number of target BESs
  ^{i = 1}
  n: Number of BINARY columns
  

[Contents][Back][Next]

[Trademarks]

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