5.5.3 Pre-reading of range indexes
You can load all range indexes into the global buffer when you start the HADB server. This is called pre-reading of range indexes. To pre-read range indexes, specify the adb_sql_rngidx_preread operand in the server definition.
- Organization of this subsection
(1) Benefit of pre-reading of range indexes
Normally, at the time of the first table search after the HADB server has started, the global buffer will not yet contain any range index data. Consequently, an I/O operation will be required to load data into the global buffer, which will degrade search performance. You can eliminate this I/O operation by pre-reading range indexes.
(2) Preventing pre-read range index data from being flushed out
Pre-read range index data is also a target of flushing from the global buffer. You can prevent such flushing from occurring by taking the following steps:
-
Store pre-read range index data that you do not want to be flushed in a dedicated data DB area.
-
Estimate accurately the number of pages in the global buffer to be allocated to the data DB area that stores the pre-read range index data that you do not want to be flushed out (that is, estimate accurately the value to be specified in the -a option of the adbbuff operand in the server definition).
For details about the formula for estimating the value to be specified in the -a option, see the explanation of the -a option of the adbbuff operand in 7.2.11 Operands and options related to global buffers (command format).