Table X (File architecture): Difference between revisions
No edit summary |
|||
(43 intermediate revisions by 8 users not shown) | |||
Line 1: | Line 1: | ||
Stores all | Stores all extensions to a Model 204 [[Record (File architecture)|record]] in a separate table from the the base records (so that [[Table B (File architecture)|Table B]] contains only the base records). | ||
Table X is enabled when you reorganize or <var>[[CREATE command: File|CREATE]]</var> a file with <var>[[XSIZE parameter|XSIZE]]</var> greater than 0. | |||
Table X is available as of Model 204 version 7.1. | |||
==Summary== | ==Summary== | ||
The Table B of a Model 204 file has a maximum number of 16.7 million record numbers (IRNs, for Internal Record Numbers). By default (<var>[[XSIZE parameter|XSIZE]]</var> = 0), these "slots" are used for both base records and extension records. | |||
By implementing Table X (<var>XSIZE</var> > 0) all defined IRNs (up to 16.7 million) in Table B are available for use by base records only, allowing the storage of the maximum possible number of records in a file regardless of the number of extensions. | |||
Table X can be significantly larger than the 16.7 million limit of Table B: the product of <code>XSIZE * <var>[[XRECPPG parameter|XRECPPG]]</var></code> can be up to 512 million extension slots. | |||
When the defined capacity of Table B or Table X is exceeded, the file is marked full or the updating transaction is backed out. | |||
To view information about memory use in Table X, use the <var>[[TABLEX command|TABLEX]]</var> command. | |||
Table X | ==Using Table X== | ||
While similar in concept to Table B, the actual "extension record number" in Table X is far more straightforward: | |||
<ul> | |||
<li>Because the number of slots in Table X is so large, and there is effectively no impact in wasting slots (until you approach that limit) <var>[[XRECPPG parameter|XRECPPG]]</var> can usually be set relatively high. </li> | |||
<li>Similarly, depending on how dynamic record growth is, you might want to set <var>[[XRESERVE parameter|XRESERVE]]</var> higher than you would normally consider setting <var>[[BRESERVE parameter|BRESERVE]]</var>. This will enable records to grow without lengthening the extension chain. </li> | |||
<li>Table X slots are always reusable. There is no special setting (equivalent to <var>[[FILEORG parameter|FILEORG]]</var> X'04') necessary to enable the use of the reuse queue. </li> | |||
</ul> | |||
==RECRDOPT parameter== | |||
Table X can be implemented in one of two ways depending on the setting of the <var>[[RECRDOPT parameter|RECRDOPT]]</var> parameter. | |||
<var>RECRDOPT</var> is set only at file creation; you cannot reset it later. To use <var>RECRDOPT</var> with files created prior to Model 204 version 7.1, you must reorganize the files under Model 204 version 7.1 or later. | |||
===The default setting of RECRDOPT (X'00')=== | |||
If the default setting of the <var>RECRDOPT</var> parameter is used, while all extensions are stored in Table X, the record allocation in Table B is identical to what it is without Table X enabled. | |||
<var>[[BRECPPG parameter|BRECPPG]]</var> remains the ''maximum number'' of records that can occur on a page, so IRNs might still be wasted (never assigned to a record) as data fills up a page. | |||
==Setting | ===Setting RECRDOPT parameter to X'01'=== | ||
This setting changes the base record allocation to ensure that all defined IRNs are available for use. | This setting changes the base record allocation to ensure that all defined IRNs are available for use. | ||
For all | For all X'01' <var>RECRDOPT</var> files, a base record size limit (the read only parameter <var>[[BRLIMSZ parameter|BRLIMSZ]]</var>) is calculated: | ||
<ul> | |||
<li>For an entry order file: | |||
<p class="code">BRLIMSZ = (6140 / BRECPPG) - 2 </p></li> | |||
<li>For an unordered file: | |||
<p class="code">BRLIMSZ = (6136 / BRECPPG) – 2 </p></li> | |||
</ul> | |||
If <code>RECRDOPT=1</code>, the maximum allowed value of <var>BRECPPG</var> is 1022 for ordered files and 1023 for unordered files. | |||
When a new record is stored, rather than using any and all space available on a page, only up to | When a new record is stored, rather than using any and all space available on a page, only up to <var>BRLIMSZ</var> bytes will be stored in Table B. This ensures that the full <var>[[BSIZE parameter|BSIZE]]</var> * BRECPPG number of IRNs can actually be stored in the file. | ||
<var>BRLIMSZ</var> must be large enough to hold all preallocated fields. | |||
(Note that the <var>BRESERVE</var> parameter is ignored (treated as a 0) with X'01' <var>RECRDOPT</var> files). | |||
==Advantages of Using Table X== | ==Advantages of Using Table X== | ||
===Scalability=== | ===Scalability=== | ||
The obvious advantage of implementing Table X relates to the ability to store the full 16.7 million base records in a Model 204 file. | |||
===Indexing performance=== | |||
===Indexing | |||
Even if a file has significantly fewer records than the limit, you should consider the performance of your indices in deciding whether to utilize Table X, notably where you expect to have fields with non unique values. | Even if a file has significantly fewer records than the limit, you should consider the performance of your indices in deciding whether to utilize Table X, notably where you expect to have fields with non unique values. | ||
When Table X is enabled and | When Table X is enabled and <code>RECRDOPT = X'01'</code>, the fact there are no wasted slots means that the bitmaps containing the bitmap and list pages (and, of course, the [[Table D (File architecture)#Existence bitmap|existence bitmap]]) are as compressed as possible. | ||
===Avoidance of | ===Avoidance of "sick record" snaps=== | ||
When large sets of unlocked records (either through the use of <var>[[Record level locking and concurrency control#FIND WITHOUT LOCKS statement|Find Without Locks]]</var> (<var>FDWOL</var>) or by referencing records on a Model 204 list) are processed, it is very possible that the records will have had their contents changed or even been deleted by the time the record is read. There are three problems that might be encountered when this happens: | |||
<ul> | |||
<li>The record should no longer be selected, its contents no longer matching the selection criteria | |||
<li>The record has been deleted (and so 'Nonexistent record' messages are produced) | |||
<li>The record now (due to delete and update processing) contains an extension record | |||
</ul> | |||
The last of these problems causes a "sick record" snap. | |||
While Rocket Software encourages the use of coding techniques to prevent all of these issues, File Managers can protect themselves from the snaps (at least) by enabling Table X in files where such processing is expected. | |||
While Rocket encourages the use of | |||
==Parameters related to the use of Table X== | |||
<table class="thJustBold"> | |||
<tr><th>[[BRLIMSZ parameter|BRLIMSZ]]</th> | |||
<td>Calculated maximum size of base records</td></tr> | |||
<tr><th>[[RECRDOPT parameter|RECRDOPT]]</th> | |||
<td>Controls the method by which base records are stored</td></tr> | |||
<tr><th>[[XAUTOINC parameter|XAUTOINC]]</th> | |||
<td>Number of pages for Table X auto increase</td></tr> | |||
<tr><th>[[XHIGHPG parameter|XHIGHPG]]</th> | |||
<td>Table X highest active page</td></tr> | |||
<tr><th>[[XQLEN parameter|XQLEN]]</th> | |||
<td>The number of Table X pages in the reuse queue</td></tr> | |||
<tr><th>[[XRECPPG parameter|XRECPPG]]</th> | |||
<td>The maximum number of extension slots on a Table X page</td></tr> | |||
<tr><th>[[XRESERVE parameter|XRESERVE]]</th> | |||
<td>Table X reserve space per page</td></tr> | |||
<tr><th>[[XREUSE parameter|XREUSE]]</th> | |||
<td>Free space required to reuse a table X page</td></tr> | |||
<tr><th>[[XREUSED parameter|XREUSED]]</th> | |||
<td>The number of extension slots reused in table X</td></tr> | |||
<tr><th>[[XSIZE parameter|XSIZE]]</th> | |||
<td>The number of pages in Table X</td></tr> | |||
</table> | |||
The use of these parameters is discussed in [[File size calculation in detail#Sizing Tables B and X|Sizing Tables B and X]]. | |||
[[Category:File | [[Category:File architecture]] | ||
Latest revision as of 22:21, 29 January 2014
Stores all extensions to a Model 204 record in a separate table from the the base records (so that Table B contains only the base records).
Table X is enabled when you reorganize or CREATE a file with XSIZE greater than 0.
Table X is available as of Model 204 version 7.1.
Summary
The Table B of a Model 204 file has a maximum number of 16.7 million record numbers (IRNs, for Internal Record Numbers). By default (XSIZE = 0), these "slots" are used for both base records and extension records.
By implementing Table X (XSIZE > 0) all defined IRNs (up to 16.7 million) in Table B are available for use by base records only, allowing the storage of the maximum possible number of records in a file regardless of the number of extensions.
Table X can be significantly larger than the 16.7 million limit of Table B: the product of XSIZE * XRECPPG
can be up to 512 million extension slots.
When the defined capacity of Table B or Table X is exceeded, the file is marked full or the updating transaction is backed out.
To view information about memory use in Table X, use the TABLEX command.
Using Table X
While similar in concept to Table B, the actual "extension record number" in Table X is far more straightforward:
- Because the number of slots in Table X is so large, and there is effectively no impact in wasting slots (until you approach that limit) XRECPPG can usually be set relatively high.
- Similarly, depending on how dynamic record growth is, you might want to set XRESERVE higher than you would normally consider setting BRESERVE. This will enable records to grow without lengthening the extension chain.
- Table X slots are always reusable. There is no special setting (equivalent to FILEORG X'04') necessary to enable the use of the reuse queue.
RECRDOPT parameter
Table X can be implemented in one of two ways depending on the setting of the RECRDOPT parameter.
RECRDOPT is set only at file creation; you cannot reset it later. To use RECRDOPT with files created prior to Model 204 version 7.1, you must reorganize the files under Model 204 version 7.1 or later.
The default setting of RECRDOPT (X'00')
If the default setting of the RECRDOPT parameter is used, while all extensions are stored in Table X, the record allocation in Table B is identical to what it is without Table X enabled.
BRECPPG remains the maximum number of records that can occur on a page, so IRNs might still be wasted (never assigned to a record) as data fills up a page.
Setting RECRDOPT parameter to X'01'
This setting changes the base record allocation to ensure that all defined IRNs are available for use.
For all X'01' RECRDOPT files, a base record size limit (the read only parameter BRLIMSZ) is calculated:
- For an entry order file:
BRLIMSZ = (6140 / BRECPPG) - 2
- For an unordered file:
BRLIMSZ = (6136 / BRECPPG) – 2
If RECRDOPT=1
, the maximum allowed value of BRECPPG is 1022 for ordered files and 1023 for unordered files.
When a new record is stored, rather than using any and all space available on a page, only up to BRLIMSZ bytes will be stored in Table B. This ensures that the full BSIZE * BRECPPG number of IRNs can actually be stored in the file.
BRLIMSZ must be large enough to hold all preallocated fields.
(Note that the BRESERVE parameter is ignored (treated as a 0) with X'01' RECRDOPT files).
Advantages of Using Table X
Scalability
The obvious advantage of implementing Table X relates to the ability to store the full 16.7 million base records in a Model 204 file.
Indexing performance
Even if a file has significantly fewer records than the limit, you should consider the performance of your indices in deciding whether to utilize Table X, notably where you expect to have fields with non unique values.
When Table X is enabled and RECRDOPT = X'01'
, the fact there are no wasted slots means that the bitmaps containing the bitmap and list pages (and, of course, the existence bitmap) are as compressed as possible.
Avoidance of "sick record" snaps
When large sets of unlocked records (either through the use of Find Without Locks (FDWOL) or by referencing records on a Model 204 list) are processed, it is very possible that the records will have had their contents changed or even been deleted by the time the record is read. There are three problems that might be encountered when this happens:
- The record should no longer be selected, its contents no longer matching the selection criteria
- The record has been deleted (and so 'Nonexistent record' messages are produced)
- The record now (due to delete and update processing) contains an extension record
The last of these problems causes a "sick record" snap.
While Rocket Software encourages the use of coding techniques to prevent all of these issues, File Managers can protect themselves from the snaps (at least) by enabling Table X in files where such processing is expected.
BRLIMSZ | Calculated maximum size of base records |
---|---|
RECRDOPT | Controls the method by which base records are stored |
XAUTOINC | Number of pages for Table X auto increase |
XHIGHPG | Table X highest active page |
XQLEN | The number of Table X pages in the reuse queue |
XRECPPG | The maximum number of extension slots on a Table X page |
XRESERVE | Table X reserve space per page |
XREUSE | Free space required to reuse a table X page |
XREUSED | The number of extension slots reused in table X |
XSIZE | The number of pages in Table X |
The use of these parameters is discussed in Sizing Tables B and X.