The ADOdb Performance Monitoring Library

V4.92a 29 Aug 2006 (c) 2000-2006 John Lim (jlim#natsoft.com.my)

This software is dual licensed using BSD-Style and LGPL. This means you can use it in compiled proprietary and commercial products.

Useful ADOdb links: Download   Other Docs

Introduction

This module, part of the ADOdb package, provides both CLI and HTML interfaces for viewing key performance indicators of your database. This is very useful because web apps such as the popular phpMyAdmin currently do not provide effective database health monitoring tools. The module provides the following:

• A quick health check of your database server using $perf->HealthCheck() or $perf->HealthCheckCLI().
• User interface for performance monitoring, $perf->UI(). This UI displays:
  • the health check,
  • all SQL logged and their query plans,
  • a list of all tables in the current database
  • an interface to continiously poll the server for key performance indicators such as CPU, Hit Ratio, Disk I/O
  • a form where you can enter and run SQL interactively.
• Gives you an API to build database monitoring tools for a server farm, for example calling $perf->DBParameter('data cache hit ratio') returns this very important statistic in a database independant manner.

ADOdb also has the ability to log all SQL executed, using LogSQL. All SQL logged can be analyzed through the performance monitor UI. In the View SQL mode, we categorize the SQL into 3 types:

• Suspicious SQL: queries with high average execution times, and are potential candidates for rewriting
• Expensive SQL: queries with high total execution times (#executions * avg execution time). Optimizing these queries will reduce your database server load.
• Invalid SQL: queries that generate errors.

Each query is hyperlinked to a description of the query plan, and every PHP script that executed that query is also shown.

Please note that the information presented is a very basic database health check, and does not provide a complete overview of database performance. Although some attempt has been made to make it work across multiple databases in the same way, it is impossible to do so. For the health check, we do try to display the following key database parameters for all drivers:

• data cache size - The amount of memory allocated to the cache.
• data cache hit ratio - A measure of how effective the cache is, as a percentage. The higher, the better.
• current connections - The number of sessions currently connected to the database.

You will need to connect to the database as an administrator to view most of the parameters.

Code improvements as very welcome, particularly adding new database parameters and automated tuning hints.

Usage

Currently, the following drivers: mysql, postgres, oci8, mssql, informix and db2 are supported. To create a new performance monitor, call NewPerfMonitor( ) as demonstrated below:

<?php
include_once('adodb.inc.php');
session_start(); # session variables required for monitoring
$conn = ADONewConnection($driver);
$conn->Connect($server,$user,$pwd,$db);
$perf =& NewPerfMonitor($conn);
$perf->UI($pollsecs=5);
?>

It is also possible to retrieve a single database parameter:

$size = $perf->DBParameter('data cache size');

Thx to Fernando Ortiz for the informix module.

Methods

function UI($pollsecs=5)

Creates a web-based user interface for performance monitoring. When you click on Poll, server statistics will be displayed every $pollsecs seconds. See Usage above.

Since 4.11, we allow users to enter and run SQL interactively via the "Run SQL" link. To disable this for security reasons, set this constant before calling $perf->UI().

define('ADODB_PERF_NO_RUN_SQL',1);

Sample output follows below:

function HealthCheck()

Returns database health check parameters as a HTML table. You will need to echo or print the output of this function,

function HealthCheckCLI()

Returns database health check parameters formatted for a command line interface. You will need to echo or print the output of this function. Sample output for mysql:

-- Ratios -- 
          MyISAM cache hit ratio => 56.5635738832 
          InnoDB cache hit ratio => 0 
             sql cache hit ratio => 0 
 -- IO -- 
                      data reads => 2622 
                     data writes => 2415.5 
 -- Data Cache -- 
          MyISAM data cache size => 512K 
             BDB data cache size => 8388600
          InnoDB data cache size => 8M
 -- Memory Pools -- 
                read buffer size => 131072 
                sort buffer size => 65528 
                     table cache => 4 
 -- Connections -- 
             current connections => 3
                 max connections => 100

function Poll($pollSecs=5)

Run in infinite loop, displaying the following information every $pollSecs. This will not work properly if output buffering is enabled. In the example below, $pollSecs=3:

Accumulating statistics...
 Time   WS-CPU%   Hit%   Sess        Reads/s          Writes/s
11:08:30    0.7  56.56      1         0.0000            0.0000
11:08:33    1.8  56.56      2         0.0000            0.0000
11:08:36   11.1  56.55      3         2.5000            0.0000
11:08:39    9.8  56.55      2         3.1121            0.0000
11:08:42    2.8  56.55      1         0.0000            0.0000
11:08:45    7.4  56.55      2         0.0000            1.5000

WS-CPU% is the Web Server CPU load of the server that PHP is running from (eg. the database client), and not the database. The Hit% is the data cache hit ratio. Sess is the current number of sessions connected to the database. If you are using persistent connections, this should not change much. The Reads/s and Writes/s are synthetic values to give the viewer a rough guide to I/O, and are not to be taken literally.

function SuspiciousSQL($numsql=10)

Returns SQL which have high average execution times as a HTML table. Each sql statement is hyperlinked to a new window which details the execution plan and the scripts that execute this SQL.

The number of statements returned is determined by $numsql. Data is taken from the adodb_logsql table, where the sql statements are logged when $connection->LogSQL(true) is enabled. The adodb_logsql table is populated using $conn->LogSQL.

For Oracle, Ixora Suspicious SQL returns a list of SQL statements that are most cache intensive as a HTML table. These are data intensive SQL statements that could benefit most from tuning.

function ExpensiveSQL($numsql=10)

Returns SQL whose total execution time (avg time * #executions) is high as a HTML table. Each sql statement is hyperlinked to a new window which details the execution plan and the scripts that execute this SQL.

The number of statements returned is determined by $numsql. Data is taken from the adodb_logsql table, where the sql statements are logged when $connection->LogSQL(true) is enabled. The adodb_logsql table is populated using $conn->LogSQL.

For Oracle, Ixora Expensive SQL returns a list of SQL statements that are taking the most CPU load when run.

function InvalidSQL($numsql=10)

Returns a list of invalid SQL as an HTML table.

Data is taken from the adodb_logsql table, where the sql statements are logged when $connection->LogSQL(true) is enabled.

function Tables($orderby=1)

Returns information on all tables in a database, with the first two fields containing the table name and table size, the remaining fields depend on the database driver. If $orderby is set to 1, it will sort by name. If $orderby is set to 2, then it will sort by table size. Some database drivers (mssql and mysql) will ignore the $orderby clause. For postgresql, the information is up-to-date since the last vacuum. Not supported currently for db2.

Raw Functions

Raw functions return values without any formatting.

function DBParameter($paramname)

Returns the value of a database parameter, such as $this->DBParameter("data cache size").

function CPULoad()

Returns the CPU load of the database client (NOT THE SERVER) as a percentage. Only works for Linux and Windows. For Windows, WMI must be available.

Format of $settings Property

To create new database parameters, you need to understand $settings. The $settings data structure is an associative array. Each element of the array defines a database parameter. The key is the name of the database parameter. If no key is defined, then it is assumed to be a section break, and the value is the name of the section break. If this is too confusing, looking at the source code will help a lot!

Each database parameter is itself an array consisting of the following elements:

0. Category code, used to group related db parameters. If the category code is 'HIDE', then the database parameter is not shown when HTML() is called.
   
1. either
   1. sql string to retrieve value, eg. "select value from v\$parameter where name='db_block_size'",
   2. array holding sql string and field to look for, e.g. array('show variables','table_cache'); optional 3rd parameter is the $rs->fields[$index] to use (otherwise $index=1), and optional 4th parameter is a constant to multiply the result with (typically 100 for percentage calculations),
   3. a string prefixed by =, then a PHP method of the class is invoked, e.g. to invoke $this->GetIndexValue(), set this array element to '=GetIndexValue',
      
2. Description of database parameter. If description begins with an =, then it is interpreted as a method call, just as in (1c) above, taking one parameter, the current value. E.g. '=GetIndexDescription' will invoke $this->GetIndexDescription($val). This is useful for generating tuning suggestions. For an example, see WarnCacheRatio().

Example from MySQL, table_cache database parameter:

'table cache' => array('CACHE',            # category code
   array("show variables", 'table_cache'), # array (type 1b)
   'Number of tables to keep open'),       # description

Example Health Check Output

db2 informix mysql mssql oci8 postgres

informix
Parameter	Val ue	Description
Ratios
data cache hit ratio	95.89
IO
data reads	1883884	Page reads
data writes	1716724	Page writes
Connections
current connections	263.0	Number of sessions

Suspicious SQL

Expensive SQL