libdb/docs/programmer_reference/am_partition.html
2011-09-13 13:44:24 -04:00

412 lines
17 KiB
HTML
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
<title>Partitioning databases</title>
<link rel="stylesheet" href="gettingStarted.css" type="text/css" />
<meta name="generator" content="DocBook XSL Stylesheets V1.73.2" />
<link rel="start" href="index.html" title="Berkeley DB Programmer's Reference Guide" />
<link rel="up" href="am.html" title="Chapter 3.  Access Method Operations" />
<link rel="prev" href="am_opensub.html" title="Opening multiple databases in a single file" />
<link rel="next" href="am_get.html" title="Retrieving records" />
</head>
<body>
<div xmlns="" class="navheader">
<div class="libver">
<p>Library Version 11.2.5.2</p>
</div>
<table width="100%" summary="Navigation header">
<tr>
<th colspan="3" align="center">Partitioning databases</th>
</tr>
<tr>
<td width="20%" align="left"><a accesskey="p" href="am_opensub.html">Prev</a> </td>
<th width="60%" align="center">Chapter 3. 
Access Method Operations
</th>
<td width="20%" align="right"> <a accesskey="n" href="am_get.html">Next</a></td>
</tr>
</table>
<hr />
</div>
<div class="sect1" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="am_partition"></a>Partitioning databases</h2>
</div>
</div>
</div>
<div class="toc">
<dl>
<dt>
<span class="sect2">
<a href="am_partition.html#am_partition_keys">Specifying partition keys</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="am_partition.html#am_partition_function">Partitioning callback</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="am_partition.html#partition_file_placement">Placing partition files</a>
</span>
</dt>
</dl>
</div>
<p>
You can improve concurrency on your database reads and writes by
splitting access to a single database into multiple databases. This
helps to avoid contention for internal database pages, as well as
allowing you to spread your databases across multiple disks,
which can help to improve disk I/O.
</p>
<div class="note" style="margin-left: 0.5in; margin-right: 0.5in;">
<h3 class="title">Note</h3>
<p>
Database partitions are not supported by the C# and Java APIs at
this time.
</p>
</div>
<p>
While you can manually do this by creating and using more than one
database for your data, DB is capable of partitioning your
database for you. When you use DB's built-in database partitioning
feature, your access to your data is performed in exactly the same way
as if you were only using one database; all the work of knowing which
database to use to access a particular record is handled for you under
the hood.
</p>
<p>
Only the BTree and Hash access methods are supported for partitioned
databases.
</p>
<p>
You indicate that you want your database to be partitioned by calling
<a href="../api_reference/C/dbset_partition.html" class="olink">DB-&gt;set_partition()</a> before opening your database the first time. You can
indicate the directory in which each partition is contained using the
<a href="../api_reference/C/dbset_partition_dirs.html" class="olink">DB-&gt;set_partition_dirs()</a> method.
</p>
<p>
Once you have partitioned a database, you cannot change your
partitioning scheme.
</p>
<p>
There are two ways to indicate what key/data pairs should go on which
partition. The first is by specifying an array of <a href="../api_reference/C/dbt.html" class="olink">DBT</a>s that indicate
the minimum key value for a given partition. The second is by providing
a callback that returns the number of the partition on which a specified
key is placed.
</p>
<div class="sect2" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="am_partition_keys"></a>Specifying partition keys</h3>
</div>
</div>
</div>
<p>
For simple cases, you can partition your database by providing
an array of <a href="../api_reference/C/dbt.html" class="olink">DBT</a>s, each element of which provides the minimum
key value to be placed on a partition. There must be one fewer
elements in this array than you have partitions. The first
element of the array indicates the minimum key value for the
second partition in your database. Key values that are less
than the first key value provided in this array are placed on
the first partition (partition 0).
</p>
<div class="note" style="margin-left: 0.5in; margin-right: 0.5in;">
<h3 class="title">Note</h3>
<p>
You can use partition keys only if you are using the Btree
access method.
</p>
</div>
<p>
For example, suppose you had a database of fruit, and you want
three partitions for your database. Then you need a <a href="../api_reference/C/dbt.html" class="olink">DBT</a> array
of size two. The first element in this array indicates the
minimum keys that should be placed on partition 1. The second
element in this array indicates the minimum key value placed on
partition 2. Keys that compare less than the first <a href="../api_reference/C/dbt.html" class="olink">DBT</a> in the
array are placed on partition 0.
</p>
<p>
All comparisons are performed according to the lexicographic
comparison used by your platform.
</p>
<p>
For example, suppose you want all fruits whose names begin
with:
</p>
<div class="itemizedlist">
<ul type="disc">
<li>
<p>
'a' - 'f' to go on partition 0
</p>
</li>
<li>
<p>
'g' - 'p' to go on partition 1
</p>
</li>
<li>
<p>
'q' - 'z' to go on partition 2.
</p>
</li>
</ul>
</div>
<p>
Then you would accomplish this with the following code
fragment:
</p>
<div class="note" style="margin-left: 0.5in; margin-right: 0.5in;">
<h3 class="title">Note</h3>
<p>
The <a href="../api_reference/C/dbset_partition.html" class="olink">DB-&gt;set_partition()</a> partition callback parameter must
be <code class="literal">NULL</code> if you are using an array of
<a href="../api_reference/C/dbt.html" class="olink">DBT</a>s to partition your database.
</p>
</div>
<a id="prog_am10"></a>
<pre class="programlisting">DB *dbp = NULL;
DB_ENV *envp = NULL;
DBT partKeys[2];
u_int32_t db_flags;
const char *file_name = "mydb.db";
int ret;
...
/* Skipping environment open to shorten this example */
/* Initialize the DB handle */
ret = db_create(&amp;dbp, envp, 0);
if (ret != 0) {
fprintf(stderr, "%s\n", db_strerror(ret));
return (EXIT_FAILURE);
}
/* Setup the partition keys */
memset(&amp;partKeys[0], 0, sizeof(DBT));
partKeys[0].data = "g";
partKeys[0].size = sizeof("g") - 1;
memset(&amp;partKeys[1], 0, sizeof(DBT));
partKeys[1].data = "q";
partKeys[1].size = sizeof("q") - 1;
dbp-&gt;set_partition(dbp, 3, partKeys, NULL);
/* Now open the database */
db_flags = DB_CREATE; /* Allow database creation */
ret = dbp-&gt;open(dbp, /* Pointer to the database */
NULL, /* Txn pointer */
file_name, /* File name */
NULL, /* Logical db name */
DB_BTREE, /* Database type (using btree) */
db_flags, /* Open flags */
0); /* File mode. Using defaults */
if (ret != 0) {
dbp-&gt;err(dbp, ret, "Database '%s' open failed",
file_name);
return (EXIT_FAILURE);
} </pre>
</div>
<div class="sect2" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="am_partition_function"></a>Partitioning callback</h3>
</div>
</div>
</div>
<p>
In some cases, a simple lexicographical comparison of key data
will not sufficiently support a partitioning scheme. For
those situations, you should write a partitioning function.
This function accepts a pointer to the <a href="../api_reference/C/db.html" class="olink">DB</a> and the <a href="../api_reference/C/dbt.html" class="olink">DBT</a>, and
it returns the number of the partition on which the key
belongs.
</p>
<p>
Note that <a href="../api_reference/C/db.html" class="olink">DB</a> actually places the key on the partition
calculated by:
</p>
<pre class="programlisting">returned_partition modulo number_of_partitions</pre>
<p>
Also, remember that if you use a partitioning function when you
create your database, then you must use the same partitioning
function every time you open that database in the future.
</p>
<p>
The following code fragment illustrates a partition callback:
</p>
<a id="prog_am11"></a>
<pre class="programlisting">u_int32_t db_partition_fn(DB *db, DBT *key) {
char *key_data;
u_int32_t ret_number;
/* Obtain your key data, unpacking it as necessary
* Here, we do the very simple thing just for illustrative purposes.
*/
key_data = (char *)key-&gt;data;
/* Here you would perform whatever comparison you require to determine
* what partition the key belongs on. If you return either 0 or the
* number of partitions in the database, the key is placed in the first
* database partition. Else, it is placed on:
*
* returned_number mod number_of_partitions
*/
ret_number = 0;
return ret_number;
} </pre>
<p>
You then cause your partition callback to be used by providing it
to the <a href="../api_reference/C/dbset_partition.html" class="olink">DB-&gt;set_partition()</a> method, as illustrated by the following
code fragment.
</p>
<div class="note" style="margin-left: 0.5in; margin-right: 0.5in;">
<h3 class="title">Note</h3>
<p>
The <a href="../api_reference/C/dbset_partition.html" class="olink">DB-&gt;set_partition()</a> <a href="../api_reference/C/dbt.html" class="olink">DBT</a> array parameter must
be <code class="literal">NULL</code> if you are using a partition
call back to partition your database.
</p>
</div>
<a id="prog_am12"></a>
<pre class="programlisting">DB *dbp = NULL;
DB_ENV *envp = NULL;
u_int32_t db_flags;
const char *file_name = "mydb.db";
int ret;
...
/* Skipping environment open to shorten this example */
/* Initialize the DB handle */
ret = db_create(&amp;dbp, envp, 0);
if (ret != 0) {
fprintf(stderr, "%s\n", db_strerror(ret));
return (EXIT_FAILURE);
}
dbp-&gt;set_partition(dbp, 3, NULL, db_partition_fn);
/* Now open the database */
db_flags = DB_CREATE; /* Allow database creation */
ret = dbp-&gt;open(dbp, /* Pointer to the database */
NULL, /* Txn pointer */
file_name, /* File name */
NULL, /* Logical db name */
DB_BTREE, /* Database type (using btree) */
db_flags, /* Open flags */
0); /* File mode. Using defaults */
if (ret != 0) {
dbp-&gt;err(dbp, ret, "Database '%s' open failed",
file_name);
return (EXIT_FAILURE);
} </pre>
</div>
<div class="sect2" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="partition_file_placement"></a>Placing partition files</h3>
</div>
</div>
</div>
<p>
When you partition a database, a database file is created on
disk in the same way as if you were not partitioning the
database. That is, this file uses the name you provide to the
<a href="../api_reference/C/dbopen.html" class="olink">DB-&gt;open()</a> <code class="literal">file</code> parameter.
</p>
<p>
However, DB then also creates a series of database files on
disk, one for each partition that you want to use. These
partition files share the same name as the database file name,
but are also number sequentially. So if you create a database
named <code class="filename">mydb.db</code>, and you create 3 partitions
for it, then you will see the following database files on disk:
</p>
<pre class="programlisting"> mydb.db
__dbp.mydb.db.000
__dbp.mydb.db.001
__dbp.mydb.db.002 </pre>
<p>
All of the database's contents go into the numbered database
files. You can cause these files to be placed in different
directories (and, hence, different disk partitions or even
disks) by using the <a href="../api_reference/C/dbset_partition_dirs.html" class="olink">DB-&gt;set_partition_dirs()</a> method.
</p>
<p>
<a href="../api_reference/C/dbset_partition_dirs.html" class="olink">DB-&gt;set_partition_dirs()</a> takes a NULL-terminated array of
strings, each one of which should represent an existing
filesystem directory.
</p>
<p>
If you are using an environment, the directories specified
using <a href="../api_reference/C/dbset_partition_dirs.html" class="olink">DB-&gt;set_partition_dirs()</a> must also be included in the
environment list specified by <a href="../api_reference/C/envadd_data_dir.html" class="olink">DB_ENV-&gt;add_data_dir()</a>.
</p>
<p>
If you are not using an environment, then the the directories
specified to <a href="../api_reference/C/dbset_partition_dirs.html" class="olink">DB-&gt;set_partition_dirs()</a> can be either complete
paths to currently existing directories, or paths relative to
the application's current working directory.
</p>
<p>
Ideally, you will provide <a href="../api_reference/C/dbset_partition_dirs.html" class="olink">DB-&gt;set_partition_dirs()</a> with an array
that is the same size as the number of partitions you are
creating for your database. Partition files are then placed
according to the order that directories are contained in the
array; partition 0 is placed in directory_array[0], partition 1
in directory_array[1], and so forth. However, if you provide an
array of directories that is smaller than the number of
database partitions, then the directories are used on a
round-robin fashion.
</p>
<p>
You must call <a href="../api_reference/C/dbset_partition_dirs.html" class="olink">DB-&gt;set_partition_dirs()</a> before you create your
database, and before you open your database each time
thereafter. The array provided to <a href="../api_reference/C/dbset_partition_dirs.html" class="olink">DB-&gt;set_partition_dirs()</a> must not
change after the database has been created.
</p>
</div>
</div>
<div class="navfooter">
<hr />
<table width="100%" summary="Navigation footer">
<tr>
<td width="40%" align="left"><a accesskey="p" href="am_opensub.html">Prev</a> </td>
<td width="20%" align="center">
<a accesskey="u" href="am.html">Up</a>
</td>
<td width="40%" align="right"> <a accesskey="n" href="am_get.html">Next</a></td>
</tr>
<tr>
<td width="40%" align="left" valign="top">Opening multiple databases in a single file </td>
<td width="20%" align="center">
<a accesskey="h" href="index.html">Home</a>
</td>
<td width="40%" align="right" valign="top"> Retrieving records</td>
</tr>
</table>
</div>
</body>
</html>