DBPLUS Performance Monitor for Oracle

Transcription

1 DBPLUS Performance Monitor for Oracle User s Manual February 2016 UM-ORA-EN-R01

2 Table of contents 1 Introduction DBPLUS Technical Support System architecture System requirements Installation of DBPLUS Performance Monitor System Configuration The main configurator screen Setting up DBPLUSORACLECATCHER monitoring service System Repository configuration The database name for the database repository Database repository parameters Login/user to connect to the base IIS service configuration User application configuration Configuration summary Adding a database for monitoring System Upgrade Setting up for the latest version License Working with program Dashboard Information bar The summary area Servers and databases area Details of ORACLE database performance Dashboard various forms of presentation Database Analysis - Performance Database Load Tab Waits Tab

3 6.2.3 Latch Tab SQL Analyze Tab SQL Details Tab SQL Plan Tab Load Trends Tab Top SQL Tab Top Day Tab I/O Stats Window Space Monitor Window Current Space Tab Current Tablespace Details Tab Overview Tab History Tab Okno Memory SGA Tab PGA Tab SGA History Tab PGA History Tab Sessions Window Backups Window Locks Window Table Locks Tab Locks history Tab Parameters Window Logs Window DBPLUS procedure statistics Tab Other Funcionalities

4 1 Introduction What is DBPLUS Performance Monitor? DBPLUS Performance Monitor tool is the software used for monitoring and analyzing the Oracle database performance Using DBPLUS Performance Monitor, you can: observe the current database performance, track trends of database server load and the individual components: CPU, I / O buffers identify performance issues of Oracle databases track performance trends of individual SQL queries analyze data and present them in graphical form watch in real time active user sessions observe the status of full and incremental databases backups troubleshoot a non-optimal SQL queries legibly report database problems and many, many more... Question: "Why do database work too slow in any specified period of time?" will never be left without an answer! 4

5 1.1 DBPLUS Technical Support Technical support provides the access to new software updates published 4 times a year as well as to engineers - help in DBPLUS Oracle database diagnosis (by using DBPLUS Performance Monitor software). 1.2 System architecture The system is designed in client-server architecture and in the presented solution we can distinguish the following components: Databases - a list of Oracle databases covered by the monitoring, Server program - an application running as a windows service, which consists of a set of procedures performed on individual databases. The aim of the program is to run periodically procedures, which are responsible for collecting basic data about Oracle servers performance. According to the DBPLUS nomenclature, program is called DBPLUSORACLECATCHER and one-up cycle within the service DBPLUSORACLECATCHER is called a snap. Repository - selected database that stores performance statistics of monitored databases. Collected statistics are the result of the work of DBPLUSORACLECATCHER service. Application - this is a client of the system, which implements user interface which allows to implement functionality of the system, i.e. monitoring review, performance analysis, query execution statistics reports, the current sessions of database, chart of server load, etc. The application is made in web technology using IIS application server and it is accessible from a web browser. DBPLUS Performance Monitor requires the installation and configuration of each of the elements to ensure full functionality of the solution. Below we present a general model of the system: 5

6 IMPORTANT: DBPLUS System Performance Monitor requires the installation and configuration on any given server / computer in the company. During normal use of application, system does not require any installation on the user's local computers. 1.3 System requirements Parameter Monitored Oracle database Server operating system with installed DBPLUS PERFORMANCE MONITOR software Description Supported types of monitored SQL database: Monitored all versions of Oracle database from version 10.2 and above Servers: Windows Server 2008 Windows Server 2012 Also: Windows 7/8 Additional requirements:.net Framework 4.0 (for DBPLUSORACLECATCHER service).net Framework 4.0 (for the client application). 6

7 Parameter The impact of the system to Oracle servers Description On the server / computer with DBPLUS Performance Monitor software is not required to install Oracle components. The system generates an average load of less than 1% dependent on generally accepted "quality" of databases As a result of the installation of repository on a selected database, the system sets up: DBPLUS scheme objects - tables o User functions with permissions that allow to read the views of systems. Warning! DBPLUS database user does not have permission to read data from the database scheme of other users. As a result of inclusion in the monitoring process a specific database it is set up the user used only to connect with a given database User interface The user application is accessible from a web browser. Supported browsers include: Internet Explorer (ver. 9 and above) Google Chrome Mozilla Firefox Opera 1.4 Installation of DBPLUS Performance Monitor DBPLUS Performance Monitor is available on DBPLUS server through the provided link. User can install DBPLUS Performance Monitor by double-clicking downloaded EXE file: 7

8 By clicking "Next" we get information about the license: In order to continue installation, you should read and accept the terms of the license. The next step is to select the directory, where DBPLUS Performance Monitor will be installed 8

9 Default directory is C:\Program Files (x86)\dbplus.oracle Visible progress of the installation process: 9

10 After the correct installation You will receive the following information The installation process is completed by pressing "Finish" button. By default, we start system configurator, which will carry out the further process of installing individual components of the system or configuration. Installed DBPLUS Performance Monitor is in the menu: Start - Programs DBPLUS The following tools are available after the correct installation 1. DBPLUS Configuration Wizard 2. Uninstall 10

11 2 System Configuration In the first stage you must set up a system on the server with DBPLUS Performance Monitor TM installed, in order to: Create a DBPLUS database repository in the selected Oracle database, which will store all the information about Oracle databases performance, Inclusion Oracle servers in the monitoring process, Configuration monitoring service DBPLUSORACLECATCHER responsible for gathering information about individual servers' performance, User Application Configuration For performing these tasks, system requires permissions: SYS or SYSTEM user or a user with DBA permissions to create the user DBPLUS and database repository After completing these steps, the application will be available to user from a web browser. 2.1 The main configurator screen On the server, where software has been installed, by clicking Start Programs DBPLUS ORACLE DBPLUS Configuration Wizard we open a window with system management tool 11

12 The main window shows the system architecture and informs among others about: number of monitored databases place in which monitoring data is stored (Database Repository), installation / configuration of individual components of system, for example: o lack of monitored Oracle databases, o user application installed or not, application services (IIS website, application pool) running or not, o if the monitoring service is enabled. In order to perform basic system configuration, click [Configuration Wizard] button and - as a result - we get this screen to configure individual components. 12

13 By default, system selects components that require configuration. You can always reconfigure e.g. a monitoring service or add another (not included so far) SQL database to monitoring. In the initial stage: We create the DBPLUS database repository We include / IIS role/service on the current machine We configure DBPLUSORACLECATCHER monitoring We configure user application 2.2 Setting up DBPLUSORACLECATCHER monitoring service DBPLUSORACLECATCHER is a program that runs as a Windows service. In the current version this service can operate using a local account. Service configuration screen is presented below: 13

14 Click on the button [Continue] to advance to the next configuration item. NOTE: All settings - made in the various components of system - are ultimately confirmed in the final step of the creator. 2.3 System Repository configuration DBPLUS Performance Monitor System Repository is a database that must be created on the selected Oracle database. To do this, you have to enter among others. Database name User name (from which the installation will be carried out) - it is mostly a system or sys - the name shall be given only once during the installation process and it is not subsequently used or stored anywhere. Login/user name that will be used to connect to a given database for this user will occur subsequent connection to database repository. This user does not have 14

15 permission to read data from schemes other than system dictionaries (owner: SYS or SYSTEM) The database name for the database repository In the first step, system asks for the basic information: The name of Oracle database - you can login using TNS or by entering the full server data without using TNS (connection type: Basic) Determining a user account with sufficient permissions to create a new user that will be used to complete the installation database repository. 15

16 2.3.2 Database repository parameters The next step is to provide information about database, such as: Name of the tablespace in which data will be kept - failure to provide the name will cause that data will be kept in the default tablespace of created / selected user The path to data file Initial file sizes 16

17 2.3.3 Login/user to connect to the base Then you have to specify the account parameters that will be used to connect to database. 17

18 2.4 IIS service configuration Launching of the IIS role / function on the server is required to run the user interface. The creator window announces additions / services of IIS application server that will be installed. If the Missing IIS features components box is empty, no configuration is required. 2.5 User application configuration Another element is the creation of user interface objects. Belong to tchem: Application s website - DBPLUS Website Application s pool - AppPoolDPMOracle Parameters: o Port number o Authentication Type when You choosing a username type = LocalService there is no need to give a user name and a password, the service will operate on default user for Windows (LocalService) 18

19 o The way to access the application - whether users at the login to the site will be asked to authenticate (login and password) or not As a result of the entire setup process completion, your application will be available at the following url: If the system will be running on port 80, link will be as follows: Click on the [Continue] button to proceed the next step 19

20 2.6 Configuration summary The last configuration step is to confirm all settings according to steps defined in the configurator. The final screen shows a summary. To confirm changes, click on [Finish] button. In the meantime, information about the progress of work is displayed 20

21 At the end of the completion status of the installation/configuration 21

22 As a result, system configuration main window looks like this: From the above sample screen, we can read that DBPLUS system PEROFRMANCE MONITOR is: installed on the server PRZEMEK-DBPLUS (link to the application in the lower right corner) all components are properly configured (the bar with information Configured successfully ) appropriate services are running: o DBPLUSORACLECATCHER - a service responsible for database monitoring o IIS, Website, App pool - which means that the application is available to the user We have 1 monitored Oracle database Information from the monitoring of all instances (currently one) are stored in the database "Dwa12" Interface / User application is available at 22

23 3 Adding a database for monitoring After the initial configuration, you can proceed to add more databases for monitoring. For this purpose, in the main System Configurator Window we click [Add Another database] button. IMPORTANT: If the [Add another database] button is not available it is the result of a license for a specified number of databases. The second option to add an instance is clicking [Configuration Wizard] button and select the component [Include / Add Oracle database to monitoring process] 23

24 As a result, we go to the wizard to add a new database. In the first place we provide basic information: Databases' names The name of database account with the privileges of the DBA, which will be used to carry out the process. 24

25 When we entering the name of the database, we can be defined connection without using TNS. The next step is to confirm the default user tablespace and the default user temp tablespace necessity to change this data should occur very rarely. 25

26 When you click [Continue] it shows a screen with the name of the user on whose will log in a service collecting data (DBPLUSORACLECATCHER), it can be existing or newly created user, during the installation user required permissions are verified (the ability to read system dictionaries) 26

27 After clicking [Continue] button, it shows the final screen being the summary of previous steps. 27

28 Click [Finish] button to add an instance to monitoring. As a result, changes are visible in the system configuration main window - DBPLUS Performance Monitor supports two ORACLE databases 28

29 When we click the link to the application (in this case we will start the application: 29

30 4 System Upgrade Technical support provides the access to new software updates that are published 4 times a year, as well as to DBPLUS engineers' help in Oracle database servers diagnosis, using DBPLUS Performance Monitor software. Upgrading system combines with two steps: Run the installation file (which goes the same as the first installation) Upgrade of database objects repository on DBPLUS user to the latest version 4.1 Setting up for the latest version In order to go through the upgrade process, you have to run DBPLUS Configuration Wizard, which also runs automatically after installation. In the result: System automatically detects the need to update to the latest version. We accept the dialog box and we run the wizard that will guide us through the upgrade process. In case of withdrawal from the operation we can always return to it by clicking [Upgrade] in the main configurator s window. 30

31 As the first screen we have information about system version, to which application will be updated (with system functionality description): The upgrade procedure applies to objects only in database where there is a DBPLUS repository. We accept by clicking [Continue] button. 31

32 System informs about operations that will perform in database on DBPLUS owner. We accept it by clicking [Finish]. Depending on version of the update process can take 1 to 3 minutes. At the end we get the information about success of the process. We close the configurator s window. 32

33 5 License The license is generated always for the server with DBPLUS Performance Monitor TM software and not for the database. Standard system license includes: System uptime The number of monitored Oracle databases Information about licensing is available from the configurator, i.e. DBPLUS Configuration Wizard After the first installation, system operates in trial version. This period lasts 14 days and system is available in its full functionality. By the end of the period it s necessary to register the system. It can be done in two ways: Sending requests for licenses from the form by clicking button [Send license request] (Internet access required on the machine) Sending computer code via 33

34 6 Working with program The user interface is accessible from a web browser at the previously configured address. The default page of the system is a dashboard showing the current performance of the monitored databases. 6.1 Dashboard After starting DBPLUS Performance Monitor TM web application, it opens a dashboard showing the current performance of monitored Oracle databases. Dashboard is divided into the following areas: Information banner summaries area servers area, instance area, Details for the selected database Information bar For information bar, user can switch a dashboard in a different mode. Available modes are: o Icon View - showing monitored servers / databases as icons (default) o Grid View - showing the databases in a table o Television Mode - showing instances of Oracle in the form of complex icons with automatically switching performance indicators 34

35 In addition, user is informed how much time is left until the next dashboard refresh by the most recent data about the current performance of all monitored instances The summary area The main area presents a general summary of: number of monitored servers and databases number of active databases Total of current waits on which databases spend their time o all waits (Summary of Waits) o disk waits (Summary of IO Waits) o Locks (Summary of Lock Waits)) From the main area you can tell whether the level of waits is high and whether to look further for the problem Servers and databases area In servers area, we see the icon of servers running ORACLE databases. After clicking on the server, in the area below these Oracle databases will illuminate which work on the machine. Icon on each server or database can be expanded by clicking on the "arrow" button or [Expand All] button 35

36 After clicking [Expand All] button at the level of database area, we see exactly which Oracle database has the highest level of waits Details of ORACLE database performance In order to analyze the current load, click on the icon of a particular database. As a result, the lower dashboard area reloads presenting details of the selected instance. Here dashboard lets to:: o observe the current CPU usage - this is a load of all processes running on the machine. o determine on what database currently spends his time Waits bookmark, Waits details bookmark o analyze performance indicators of the last 24 hours - Performance Counters o check the level of the session / locks, active transactions - Sessions o to see the load of the database from the last 24 hours - Database Load o verify memory utilization - Memory o check database size by tablespaces o view basic information about database o check alerts Information about CPU utilization, Waits, sessions are presented here in the horizon of the last 15 minutes. For example, on Waits chart, we have provided the following information: I/O waits - readings of disk devices 36

37 Locks - locks of database records by session Latches - waits for access to database buffers Chart means that at a certain point in time (the time read from the X-axis), all users (active sessions) are waiting for the outcome of the query indicated number of seconds (the result read from the Y axis). Categories I/O, Locks, Latches help to state why sessions are staying idle. By default, all series are visible: When you click in the series [Waits, IO Waits, Latches], it remains visible only [Locks] waits series (another click on the legend bar, displays selected series again) In the example, we see that Oracle database has performance issues casued by the locks. Between 8:55:45 and 9:07:30 hours locks remained at 6 seconds level (6 blocked sessions), then fell to 2 seconds. IMPORTANT: the level of waits is re-calculated to one second. 37

38 Another information about waits can be known in the next tab [Waits details]. On the dashboard screen for the selected database, you can also check the load, system indicators - so called performance counters and memory utilization of the last 24 hours. To do this, click on the appropriate tab on the left side. In the Performance Counter we should be aware of the additional selection filter of the group: The load from the last 24 hours we get after clicking on [Database Load] button. 38

39 Database Load is one of the core modules used by DBPLUS engineers to analyze performance. Chart consists of the following series: Elapsed Time - shows the summary time all users are waiting on the result of a query at a given second of time. On the graph for the displayed point Elapsed Time is 2.93 seconds, which can be interpreted as follows: o 3 users launched different queries - 2 users waited for one second, third waited 0.93 second. CPU Time the utilization of virtual server processors by all queries in a given second in time. IO Lock, Wait Time - the result of the difference between the Elapsed Time and CPU time of database Server CPU - CPU load on the server For better readability of the graph: you can click to disable (or enable) given series of the chart - we do it in the area of the chart legend you can zoom the chart Here is an example showing series of Elapsed Time and CPU Time in the narrower time horizon: Dashboard also allows you to view basic information about database, among others: database version the number of available virtual processors date of last reboot recent changes to database parameters 39

40 We get it after clicking on the [Info] tab: When you click on [Memory], you have information about memory utilization on database. In addition, we see what level of memory usage is kept by individual areas (DB cache size, Shared, Large, Java Pools): An additional feature of the dashboard is alerting about more or less critical performance events on server, among others: Increased CPU utilization on server Locks The decrease in Buffer Cache Hit Ratio The increase in the number of sessions or open transactions Etc. When you click in the [Database space], you can get to know the current size of database broken down by spaces (size is expressed in MB): 40

41 6.1.5 Dashboard various forms of presentation Dashboard is available in 3 modes, which are switched by clicking [Toggle View] icon in the top right corner. Available modes are: Icon View - showing monitored servers / databases in the form of icons (default) 41

42 Grid View - showing instances in the form of a table Television Mode - showing instances of Oracle in the form of icons with automatically switching performance indicators 42

43 Changes to the alerting and indicators appearing in the [Television Mode] can be made at the option Configuration -> Settings or clicking on the icon on the blue title bar 43

44 6.2 Database Analysis - Performance Dashboard of the DBPLUS System Performance Monitor allows you to track the performance of Oracle databases, and show how it looked over the last 15 minutes or the last 24 hours. For a detailed analysis of the load at any given moment in time, and seek answers to questions like: why database is running slow? why user had problems in the application 3 days ago at 15:48? why my report performed 15 minutes? etc.... We enter the module [Database Analysis] and here we have two possibilities: On the left side of the menu, click on [Database Analysis] shows a list of ORACLE databases Displaying the details of database after an earlier being selected on the dashboard page 44

45 6.2.1 Database Load Tab Database Load is a screen showing database load time, which was partially discussed in the previous section on the Dashboard. In the Performance module functionality of the graph is greater. First of all, we can here: check the load of the database in the wider horizon e.g. today, yesterday, a month or even 3.5 years ago. look at the SQL queries / commands, which generated the load asses what database did at this time among others if performed a lot of disk operations, whether there were locks, etc. Database Load screen consists of following areas: filtration fields - fields of dates by which we define the period in which we want to check- the load the graph presenting the load the load information at a given moment of time: o list of queries with execution statistics o Waits - what database was doing at the time to perform queries Chart consists of following series: Elapsed Time - shows the summary time all users are waiting on the result of a query at a given second of time CPU Time the utilization of virtual server processors by all queries in a given second in time. IO Lock, Wait Time - the result of the difference between the Elapsed Time and CPU time of database Server CPU - CPU load on the server The chart is "clickable" - click on the selected part / section will refresh the bottom of the screen with information about requests and waits that generated the data load. IMPORTANT - the screen shown that Oracle database is on a server with 8 CPUs. In the current day, until 8:42, there was no utilization of the database. Then there was an 45

46 increase to 1 CPU load, with a significant increase in waiting time for the users for query result (Elapsed Time). IMPORTANT - data for chart of the database load is calculated by monitoring service DBPLUSORACLECATCHER - a component of the DBPLUS Performance Monitor. Monitoring Service performs a number of procedures examining database performance. The result of the operation of these procedures is a snapshot (snap) that is created every 15 minutes. If one session for 15 minutes performed 3 queries: The first query 5 minutes The second query 09 minutes and 59 seconds The third query 1 second the graph for a given situation will present utilization of one CPU with Elapsed Time close to 1 second. After clicking on the selected point in time, the lower part of the screen is refreshed, information requests and waits If you scroll down the screen: 46

47 There are 2 additional sub-tabs: SQL Statements Waits SQL Statements is the query statistics presented in the form of a table. By default, system displays the most aggravating question for the duration of Elapsed Time or CPU utilization. The method of display can be changed after clicking on the field [Statement filter] - you can also view a complete list of queries that participated in the load. The table with queries: You can sort on any column Go to the next page- buttons Previous, Next Change the number of queries per page Search e.g. after a part of the query text The table contains the following information: SQL Text - full text of SQL command Query Hash/Query Plan Hash - an identifier of a query and an identifier of execution plan Time per one Executions (sec) duration of query in seconds for one execution, Time of duration (sec) - the duration in seconds for all query executions within last 15 minutes. 47

48 CPU time (sec) - CPU utilization time in seconds by the query during last 15 minutes. Executions - number of executions of the query in last 15 minutes. Sorts - number of sorts performed by the query during last 15 minutes Fetches Disk Reads - number of disk reads for the query in last 15 minutes, Buffer Gets - number of buffers utilized by the query during last 15 minutes Rows processed - number of rows returned by the query in last 15 minutes Percent of load by Elapsed Time - the percentage of total database load caused by the query during last 15 minutes. Percent of load by CPU Time - the percentage of total database servers CPUs load caused by the query during last 15 minutes, In the column Query Hash / Query Plan Hash (each line presenting statistics of the execution) shows [+] ("plus") button When you click on [+] ("plus"), it shows additional context menu, which enables for detailed analysis of a particular query, which will be discussed in the section "Performance SQL Details 48

49 For example, if you select "Add to query hash list", we move a query identifier to the clipboard with a list of queries for later analysis of specific queries Below the slide of 4 queries added to the analysis in SQL Details functionality. Below the table with a list of queries is presented the full text of the query with the execution plan. Clicking on particular query will refresh these fields. Under the implementation plan there is available a link to functionality enabling its analysis and analysis of objects involved in the query, among others: what tables, indexes participated in the execution of the query how the engine called the given objects search data (seek) read complete data (index or table scan) whether the query was carried out in multi-threaded mode what mechanism was used to download and connect data from objects: Nested Loop Hash/Merge Join Join When you click on a link [Show Plan Object], you get: 49

50 In Window we have information about the query text and the execution plan. Below there are areas: Objects Used in Explain Plan a list of all objects used by the query in the execution plan Indexes for selected object list of indexes for selected table - row selected in the "Objects Used in Explain Plan The area consisting of 1/3 bookmarks a. Object Columns a list of the individual columns of the selected object, along with information such as: column name, data type, column id, density (the lower density, the higher selectivity of the column) b. Info basic information about selected object, DDL code, c. Properties additional properties of selected object When we analyzing the execution plan, we pay particular attention to: Limiting the choice of data, or of the data with the where clause and tables joins Whether the request is with parameters or literals Actions that SQL Engine chose to data download Whether in the table are appropriate indexes Way of reading the data - Nested Loops vs. Hash Join Another tab (next to [SQL Statements]) is [Waits]. Here we have waits presented in graphical and tabular form. The graph shows the duration for each second of the selected snapshot (of 15 minutes) of each kind of wait that occurred at that time on the instance 50

51 Below the graph there is a table where: Wait Name - the name of wait Description - the meaning of one wait Wait time - per 1 second (sec.) - Duration of wait type in seconds Wait time - the total for snap (sec.) - the total duration of wait type in the snapshot (15 minutes) Waits Tab [Waits] tab shows the duration of waits, which occurred at a time for all sessions on ORACLE database. Depending on the selection of [Toggle View] option, data can be shown graphically or in tabular form. [Waits] screen as [Database Load] consists of the following fields: filtration fields - fields of dates by which we define the period in which we see database' waits graph presenting the level of waits detailed information about waits in a given moment of time The Y axis of the graph illustrates (in seconds for a given second) time of all waits that occurred during the period shown on the X-axis. X-axis of the graph shows the period of time in which waits occur 51

52 After switching the view with [Toggle view] button we get: Just like in the screen [Database Load], [Waits] chart is "clickable". Clicking on the part of the graph (its point) will show us waits summary, appropriate for a particular snapshot in time. 52

53 From the above chart we can see: What was database doing during the day (the default) or a selected period limited by dates in the filter what was database doing in a given moment of time The system also allows to analyze individual Waits - the frequency, length and time of occurrence. To do this, click on the sub-tab [Analyze]: Here user has the ability to analyze performance waits or all of waits grouped by day, hour, snap(shot). An example results of analysis we present below 53

54 Data are transferred on the chart area by clicking on the selected wait's row or dragging it on the chart's area Latch Tab The "Latch" tab represents time of occurrence of all LATCHES in seconds, which occurred at a given time for all user' sessions of ORACLE database. 54

55 Latch screen maintains almost the same functionality as in Waits (click-through of the chart, the presentation of all latches and latches at a specific point in time) The Y axis graph shows the time in seconds of all LATCHES that occurred at the time shown on the X-axis X-axis graph shows the time in which there are LATCHES. The chart is active and refreshes every 15 minutes by clicking the [Refresh] button. In addition, on the [Latch] screen we have the following functionality: Buffer Latches - analysis of queries that most utilize memory area Row Cache Objects - latches broken down by operations related to SHARED_POOL Latch Library Cache - allows you to check how much space queries occupy in the memory buffer SHARED_POOL 55

56 For example, the following analysis Row Cache object in one-week horizon (7 days in grouping by day): Clicking on a specific statistic: [Count], [Usage], [Gets] refreshes the graph. As the Database Load Waits, Latch, charts this chart is also "clickable" - clicking on the point will refresh the following charts below. The "Latch library cache" tab allows you to check how much space queries occupy in the memory buffer SHARED_POOL 56

57 Option [Shared Pool Statements length] allows to set the length of the query. Setting the length of the query to determine how many bytes in the SHARED_POOL buffer occupy the same queries with a precision of length of query. Queries are grouped if they are the same as the predetermined length. This is particularly useful for checking how much space occupy in the SHARED_POOL buffer queries with literals instead of variables. We can just set the length of a query to a value which ensures that the queries are up to a given length the same. Then for such queries, to the desired length, will be presented following statistics: The table contains following information: Partial SQL text - Text of the query to the desired length (in the above example, 24 characters) Memory usage (bytes) - the amount of memory used in bytes in the SHARED_POOL buffer Count - the number of existing versions of the query in the SHARED_POOL buffer 57

58 Check Box Show Statement Version enables or disables showing all versions of the queries searched for a given length of queries contained in SHARED_POOL. Depending on whether the check box "Show Statement Version" is selected or not there are shown two types of data. Statements - (check box "Show Statement Version" is selected) shows the full text of queries that are included in the predetermined length Growing Statement - (check box "Show Statement Version" is not selected) is displayed: Statement query text up to the set length Difference in size between snaps (bytes) - The difference in the size of the occupancy of SHARED_POOL buffer for all queries that have changed the occupancy between refreshes of the entire screen Statement History area is divided into two parts 58

59 The first part shows a list of changes of the size in bytes of memory used in SHARED_POOL and the number of versions of the query existing in the buffer. The second part is a graph which graphically shows the size of used memory, where: The X-axis represents the time at which the query caused SHARED POOL cache utilization. The Y-axis shows the size of the SHARED POOL buffer, in bytes used by the query. After you select the check box "Enable Auto Shared Pool History" area "Statement History" refreshes automatically every 30 seconds SQL Analyze Tab [SQL Analyze] functionality presents an additional view of database load. As with the [Database Load] chart, here we have too shown a graph with utilization of the database. The screen consists of the following areas: filtration fields: o date fields by which we define the period in which we want to refer to the instance load o way of the presentation of the load - by CPU Time or Elapsed Time the graph presenting the CPU load or Elapsed Time (depending on the choice in the filter) cumulative statistics broken down by queries, that during given period of time generated certain load 59

60 The Y-axis graph shows the number of seconds for each second of duration of the query in ORACLE database. The X-axis represents the time at which the query caused the utilization of database server. The differences that can be shown between load being shown in the Database Load chart and utilization statistics of the database server from the operating system arise due to including in the chart of all kinds of waits, which is not shown in the operating system. In the graph is shown full picture of performance and not only the time. After clicking on the 2 queries, we see their share of database load and when it took place with an accuracy to 15 minutes: 60

61 In the presented case 3 queries generated 80-90% of database load. The table in the SQL Statements tab shows statistics for each query: Query Hash/Query Plan HashValue ID of the query and the execution plan Elapsed Time (sec) - total time in seconds of query duration for all query executions taking place in the selected time period. If the filter is marked "Group by Plan," the Time of duration (sec) indicates the duration of all searches for the PLAN HASH VALUE for a selected period of time. CPU time (sec) total time in seconds of CPU utilization for all executions of the query taking place in the selected time interval. If the filter is marked "Group by Plan," time of duration (sec) indicates the duration of all searches for the HASH VALUE PLAN for a selected period of time. Sorts - The number of sorts performed by the query in the selected time Fetches - The number of returned records Executions - The number of query executions in the selected time Disk Reads - The number of disk reads for a query in the selected time Buffet Gets - The number of buffers utilized for a query in the selected time Rows processed - Number of rows returned by the query in the selected time IMPORTANT - SQL Analyze screen maintains similar functionality as in the Database Load Clicking on a query row (in addition to showing the load on the graph) will display the full text of the query and its execution plan 61

62 Next to the query identifier is the [+] ("plus") button, which adds a query to the clipboard with a list of queries SQL Details Tab [SQL Details] tab shows detailed information about the query such as: frequency, execution time, whether the request has changed execution plan, the number of returned records, the number of executions, etc. This information provides the opportunity to decide whether it makes sense to optimize given query. 62

63 [SQL Details] tab is divided into several areas: Clipboard with a list of query IDs (expanded and collapsed by clicking the green button in the lower right corner of the screen) - the queries to the clipboard are added from screens Performance -> Database Load Performance -> SQL Analyze Performance -> Top SQL Performance -> Top Day Sessions Filters area and the way to display statistics for: o specified Query Hash ID of the query o selected date range o a grouping of statistics by day, snapshot, etc. The area with the text of the query 63

64 Detailed execution statistics in form of the table Execution plan (with the selected [Explain Plan] tab) Graphic presentation (with the selected [SQL Statement Loads] tab) any indicator/column from the statistic table 64

65 Typing a query ID into the field: [Enter Query Hash] displays statistics for given query identifier according to set filters. IMPORTANT: If you do not know and do not have any query ID and clipboard with a list of queriesis empty, you can: go to one of the screens (Database Load, SQL Analyze, Top SQL, Top Day), where you can search suboptimal/long-lasting query Click on the Find SQL to search for a particular query (search by its text) Statistics show: Query Plan Hash- query plan identifier Elapsed Time (sec) - total time in seconds for the duration of the query for the selected period grouping. CPU time (sec) - total time in seconds for CPU utilization for the query for the selected period grouping. Rows processed - Number of rows processed by query for the selected period grouping Executions - The number of query executions for the selected period grouping Disk Reads - The number of disk reads for a query for the selected period grouping Buffet Gets - The number of buffers utilized by the query for a selected period grouping Elapsed time per 1 exec - duration of a single query execution for the selected period grouping Filter [Group by period] - shows statistics for a given query grouped according to the choice for: None - no grouping i.e. selection date range from 1 to 20 day of the month will show aggregate statistics for the selected period Snap - shows statistics for a given query, broken down by snapshots - periods of 15 minutes Hour - shows statistics for a given query, broken down into periods of one hour Day - shows statistics for a given query, broken down into periods of one day [Show Online Values] option It allows you to display current information about queries stored in the database buffer. 65

66 This information in other tabs will appear after 15 minutes and when you check [Show Online Values] option you can see them right away if the starts up To check whether the query is running or not at the moment, we have to, after giving the Query Hash ID, select the check box [Show Online Values] and clicking on the button [Refresh] to watch the values in the Execution, Elapsed Time, CPU Time columns. If the values are changing it means that the query still performs. If the values are constant this inquiry currently ceased to perform. By clicking on the [SQL Statement Load] tab we can see the load generated by the query (line/area yellow) on a total load of base: [Explain Plan] tab It shows the query execution plan. If there is more than one, then you can click the check box [Compare plans], which will display two execution plans. It makes it easy to find differences between them. 66

67 In the first from the top execution plan there is an active link [Show Plan Objects], which shows the objects used in the query plan: In the [Show Plan Objects] you have repeated the information about the query text and the execution plan: Under the text and the execution plan there are areas: Objects Used in Explain Plan - a list of all objects used by the query in given execution plan Indexes for selected object - list of indexes for selected table - row selected in the "Objects Used in Explain Plan" The area comprised of three tabs a. Object Columns - a list of individual columns of the selected object, along with information such as: column name, data type, id columns, density (the lower density - the higher selectivity of the column) b. Info - basic information about selected object c. Properties - additional properties of selected object When we analyzing the execution plan, we pay particular attention to: Limiting the choice of data, or of the data with the where clause and tables joins Whether the request is with parameters or literals Actions that SQL Engine chose to data download Whether in the table are appropriate indexes Way of reading the data - Nested Loops vs. Hash Join In this example, we evidently see when after change of the execution plan, the query sped up more than 40 times. Elapsed time of a single execution changed from to seconds. 67

68 [Find SQL] button In a situation where we do not know query ID, and we want to find among others: Queries that refer to a particular table Queries that changed execution plan Queries that use specific database objects e.g. an index we click on [Find SQL] button in effect we see a search query window. Below is a sample screen with (found) queries making reference to the object m_invoice: Found queries we can "move" to clipboard of [SQL Details] screen by clicking [+] ("plus") icon next to query ID SQL Plan Tab [SQL Plan] tab allows you to get detailed information about the queries using specific query plan 68

69 [SQL plan] window is divided into areas: Clipboard with execution plans IDs [SQL Plan Statistics] presents the following information: 69

70 Date - date of query execution. Hash Value query ID (if selected option Group by Query) Elapsed Time (sec) - total time in seconds of the duration of the query for the selected grouping period. CPU time (sec) - total time in seconds of CPU utilization for the query for the selected grouping period. Executions - The number of query executions for the selected grouping period Disk Reads - The number of disk reads for a query for the selected grouping period Buffet Gets - The number of buffers utilized for the query for the selected grouping period Elapsed time per 1 exec - duration of a single query execution for the selected grouping period Explain Plan It shows SQL query execution plan. Statements Using This Plan The table shows queries that use the given execution plan, which allows for a more complete assessment of the situation, it often turns out that the same plan is used by a very similar query. Statement Text Shows identifiers and the contents of all SQL queries that use particular execution plan. 70

71 6.2.7 Load Trends Tab [Load Trends] tab allows you to get detailed information on trends in ORACLE database. The window consists of three components: Filter with the date range and grouping option Graph presenting certain indicators over time The table of statistics Load Trends Statistics include the following information: Log date - represents the point in time for which the statistics are presented (i.e.: day, hour, minutes, for the entire period) Elapsed Time (sec) - total length of time in seconds of all queries for the selected grouping period. Executions - The number of performances of all searches for the selected grouping period Disk Reads - The number of disk reads for all queries for the selected grouping period Buffer Gets - The number of utilized buffers for all queries for the selected grouping period Rows processed - Number of rows processed by all queries for the selected grouping period Latches - total time in seconds the duration of all latches that occurred for the selected grouping period Waits - total time in seconds spent on all waits including latches that occurred for the selected grouping period Locks - total time in seconds of the duration of all locks that have occurred for the grouping period 71

72 After clicking on the selected columns (or dragging them to the chart area) present their behavior as a function of time: When you change the chart type for Area, graphs look like this: Top SQL Tab [Top SQL] tab allows you to see queries that create the most load for the database, depending on whether we are interested in execution time, the quantity of data read, the number of processed blocks from memory (Buffer Gets), or the number of reads from disk (Disk Reads) etc. Queries are presented in the form of several charts in descending order according to duration of the query in the given period of time (or any other selected indicator). 72

73 From the level of [Top SQL] screen you can easily add interesting query to [SQL Details] by clicking on the button "Plus" next to the query ID and clicking on options: SQL View details - to move to the SQL Details screen and analyze specific inquiry Add to query hash list - to add the query to the clipboard with a list of questions for further analysis By using check boxes in the legend we can delete individual charts from the [Top SQL] view On the right side we have a choice opportunity to present top queries by a specified indicator: Elapsed time CPU Time Sorts Fetches 73

74 Disk Reads Buffer Gets Rows Process Executions When you click the link [Show Additional filters] we have additional options, among others: Resizing the size charts for presented queries View statistics by execution plans Top Day Tab [Top Day] window allows you to view top queries for CPU Time or Elapsed Time and track of their behavior changes 74

75 In the above slide there are presented top queries during the last week and share of the first question against the load of the entire database. Conclusion: optimizing the selected query(ies) we ll reduce database load by 100%!!! Table with top queries contains Date - date of query execution Query Hash SQL query identifier - Elapsed Time (sec.) Total time of execution of SQL query CPU Time - total time of CPU usage Load Percent - percent load of the database Top Place For Day - Number of Explain of Plans - the number of execution plans Below the table we have Statement Text text of the selected query. By checking the query in the table we can drag query to chart [Database Load] and observe changes of its influence on the overall load of the database. You should remember about the possibility of detailed analysis of a particular query by clicking on the button [+] ("plus") next to the query. 75

76 6.3 I/O Stats Window The screen is accessed from the menu on the left side and is used to analyze the performance of disk components. [I/O Analyze] functionality allows you to see any performance problems on disk devices, among others comparing the performance of writing and reading for particular days, hours, data files, tablespaces as well as collectively for the entire database [I/O Stats] consists of the following features: I/O Analyze Archived Log Log file switch Days Compare, Period Compare - functionality that allows to compare IO operations at specified days or periods of time. The [I/O Analyze] allows you to see any performance problems on disk devices. The window is divided into the following sections: Filters area with the range of dates and additional filters Chart for presentation of specific indicators Table showing stats of o Reads - the number of reads o Writes - the number of data writes by DBWR o Block Reads - the number of read blocks o Block Writes - number of wrote blocks 76

77 o Read Time - time to read blocks o Write Time - time to write blocks o Block Read Time - time to read a single block o Block Write Time - time to write a single block Group by period - allows you to choose the period by which the grouping will occur Day - result is grouped by day Hour - result is grouped by hour Snap - result is grouped by periods of 15 minutes None - grouping is switched off, will show the total sum for a selected period for database, data files or table space depending on which filter is used Additional filters are available by clicking on [Show additional filters] link, among others: Grouping by space or data file The filter on the data file The filter on the tablespace That allow you to analyze I/O in a specific file or data space. [Archive Logs] tab shows the number of archive files broken down by days: The window is divided into two parts: The table that shows the statistics: Date - the date for which statistics are shown 77

78 Number of Archive Logs - the number of archive files, which was created on that day Size of Archived Logs (bytes) - the sum in bytes of all archive files that were created on that day Graph "Number of Archived Logs Group by Day" Hovering the mouse over a graph s bar allows you to see details showing how many and what size archive files were generated. The Y axis graph shows the number of created archive files. The X axis shows the time of creation of the files with an accuracy of one day The [Log file switch] tab presents in a graphical form switching of redo files. The X-axis represents the time in which there has been switching of the files. The shorter the time between "dots" on the graph, the faster were switching of redo files. The Y-axis shows the file size in MB Under the graph in the form of a table it shows data presented in the chart 78

79 6.4 Space Monitor Window [Space Monitor] module allows to analyze the space used by database. Within the tabs we have the possibility to: Display the current size of database Detailed information on database occupancy (by database objects) The history of database size change in tabular and graphical form IMPORTANT: [Space Monitor] module is also accessible from the main page (we go there after clicking [Back to dashboard]) and allows to analyze the space used by all monitored databases Current Space Tab [Current Space] tab shows the current size of database, by default expressed in GB. In the filter we can change among others way of grouping and set a filter on a specific area: 79

80 In the table below the graph, we see: Database Size The space used Statistics about size increments Current Tablespace Details Tab In [Current Tablespace Details] window user can check the size of each database object. The window consists of two tables: Object size by type - presents data grouped by type of object Object size - list of individual objects with their size 80

81 6.4.3 Overview Tab In the [Overview] screen system shows the size of database each day in the form of a table. Filter area allows us to analyze space: In the chosen time horizon For all or a specified space of database For certain types of files At a certain grouping History Tab [History] tab shows the size (in GB) of selected tablespace for each day of the selected date range. The chart area is refreshed after clicking on indicated line in the table below (in the case presented above, there are data files for USERS' space). 81

82 6.5 Okno Memory Memory module allows to analyze the utilization of memory in database. In the tabs we have the possibility to: View the current use of memory History of memory usage over time Memory usage by user sessions SGA Tab [SGA] tab shows settings for each parameter of Oracle Database SGA included in database. Parameter settings are presented as components of the wheel. The size of the individual parameters is expressed in megabytes. Clicking on an interesting piece of wheel showing the parameter, will show statistics for this parameter using the chart on the right side of the SGA wheel. Chart of statistics has different characteristics for each parameter. [SGA] window is divided into two parts: A wheel chart showing the proportion of: Shared Pool Size Java Pool Size Sga Not Used Size DB Cache Size Log Buffer Size A bar graph showing individual components where: The color of the selected area indicates quantity of the used buffer Green color states amount of free part of the buffer 82

83 The exception is the "DB Cache", where the bar graph shows the percent hit ratio in this buffer PGA Tab The [PGA] is divided into three parts: A bar graph showing the amount of memory that "PGA" occupies A bar graph showing the hit ratio in the "PGA" The table shows the utilization of the PGA buffer by each of the session, wherein: Session SID user session ID BYTES number of occupied bytes in PGA buffer OSUSER name of operating system user logged into Oracle database USERNAME Oracle database user name MACHINE name of the machine from which session logged into Oracle database STATUS user session status ACTIVE, INACTIVE, KILLED 83

84 PROGRAM name of te application from which session logged into Oracle database SGA History Tab [SGA History] tab is divided into 6 parts and shows the value of individual buffers and their utilization for a specified period of time: Buffer Hit Ratio The graph shows the cache hit ratio for DB_CACHE_SIZE buffer as well as DB_CACHE_SIZE buffer size for the selected period of time in function Group by. The Y axis on the left side of the graph shows the buffer hit ratio for DB_CACHE_SIZE buffer. The Y-axis of the graph on the right shows the size in MB of DB_CACHE_SIZE buffer. The X axis determines the time for which data are shown in the chart. SGA MAX The graph shows the value of "SGA_MAX_SIZE" in a given period of time. The Y-axis represents the value expressed in megabytes. The X axis determines the time for which data are shown in the chart with accuracy chosen by the administrator: date, time, snap (every 15 minutes). Log Buffer The graph shows the log_buffer buffer size as well as the highest number of attempts to write to the buffer because of its occupancy. The Y axis on the left side of the chart shows the log_buffer buffer size in MB. The axis Y of the right side of the graph shows the largest number of attempts to write to the log_buffer buffer due to its occupancy. 84

85 The X axis determines the time for which data is shown in the chart. Shared Pool Shared Pool The graph shows the utilization of the SHARED POOL buffer, expressed in megabytes. The Y-axis shows the value of utilization, expressed in MB. The X axis graph shows the time in which there was a utilization with accuracy chosen by the administrator so: date, time, snap (every 15 minutes). Grey color shows used space in the SHARED POOL buffer. Green color shows the free space in the SHARED POOL buffer. Java Pool The graph shows the Java Pool buffer utilization expressed in megabytes. The Y-axis graph shows the value of utilization, expressed in MB. The X axis graph shows the time in which there was a utilization with accuracy chosen by the administrator so: date, time, snap (every 15 minutes). Grey shows used space in the JAVA POOL buffer. Green color shows the free space in the JAVA POOL buffer. Large Pool The graph shows the utilization of the LARGE POOL buffer expressed in megabytes. The Y-axis graph shows the value of utilization, expressed in MB. The X axis graph shows the time in which there was a utilization with accuracy chosen by the administrator so: date, time, snap (every 15 minutes). Grey color shows used space in the LARGE POOL buffer. Green color shows the free space in the LARGE POOL buffer. Using the drop down menu "Group by", relevant charts show data grouped according to the choice: Snap - broken down by periods of 15 minutes Hour - broken down by periods of one hour Day - broken down by periods of one day 85

86 6.5.4 PGA History Tab [PGA History] Window is divided into two parts: PGA MEMORY The graph shows the utilization of the PGA buffer expressed in megabytes. The Y-axis graph shows the value of utilization, expressed in MB. The X axis graph shows the time in which there was a utilization with accuracy chosen by the administrator so: date, time, snap (every 15 minutes). Grey color shows used space in the PGA buffer. Green color shows the free space in the PGA buffer. PGA Hit Ratio The graph shows the cache hit ratio in PGA buffer expressed in %. The Y-axis graph shows the value in % cache hit by Oracle database sessions. The X axis graph shows the time with accuracy chosen by the administrator so: date, time, snap (every 15 minutes). Using the drop down menu "Group by". relevant charts show data grouped according to the choice: Snap - broken down by periods of 15 minutes Hour - broken down by periods of one hour Day - broken down by periods of one day 86

87 6.6 Sessions Window [Sessions] tab shows information about sessions in database. In the tabular part are presented following details: Logon Time - The time when the user is logged into Oracle database Sid user's session ID along with Serial# Serial# - user's session ID along with SID Hash_value - identifier of currently* executed command. * Indicates that the command is currently being executed with accuracy given by Oracle database. User name the name of Oracle database user Status - the status of the session: ACTIVE, INACTIVE, KILLED Elapsed Time - the duration of the query or block of PL/SQL in seconds Schema Name - the name of the user who executes the SQL command Os user - User name in the operating system from which the user logged to Oracle database Process Number - the number of process in operating system that handles a session Machine the name of the machine from which user is logged into Oracle database The program - the name of the program from which user is logged into Oracle database Session Wait - the name of type wait of the session [SQL] tab represents the text of SQL query. [Operation Progress] Tab shows the execution state of the current operation by a session in Oracle database [Statistics] Tab shows us information about statistics of a particular session 87

88 Session Waits Tab Session History Tab It shows us in a graph the number of sessions active and inactive during the given period of time. 88

89 Clicking on sessions at the main table will refresh information in the tabs below. 89

90 6.7 Backups Window Backups window allows you to check the performance of backups, i.e. the execution time and the history of executed backups. It is divided into three parts: A chart that shows finished backup copies of database (data files, control files, archive files, files init) that were executed in the selected date range. The Y-axis graph shows the performance of the backup copies in bytes. The X axis graph shows the time in which the backup was executed. "Output" table contains information about output files created on media on which the backup using RMAN tool has written data: Filename - the name of the output file Open time - the time to start execution of writing to media to the output file Close time - the time to stop execution of writing to media to the output file Elapsed time the time of writing to media to the output file Bytes - the size in bytes that has been written to the output file Bytes per sec. - Productivity expressed in bytes per second of write speeds to the output file Device type - the type of device on which the write was made to the output file (a tape, a disk) Type the type of operation. Output means write. Status - the status of the performed operation "Input" table contains information about files, which have undergone a backup using RMAN: Filename - the name of the file, it can be a data file of Oracle database, archive file, spfile file Open time the time to start the read of the file from the disk 90

91 Close time the time to stop the read of the file from the disk Elapsed time the duration time of the read of the file from the disk Bytes - the size in bytes that has been read for a given file Bytes per sec. - Productivity expressed in bytes per second of read speed for the given file Device type - the type of device from which a file was read Type the type of operation. INPUT means read. Status - the status of the performed operation The option All Output / All Input and clicking in the safety backup of our interest on the graph shows in two tables all files which have undergone a backup, as well it shows all the output files or file names recorded on the media. 91

92 6.8 Locks Window [Locks] module consists of the following tabs: Table Locks - allowing the current analysis of locks on database Locks history - allows tracking of locks in time Locked Objects - presenting a list of objects which are locked at the moment Library Locks Table Locks Tab [Table Locks] tab consists of the following areas: Filter bar The tree of blocked sessions: at the top of the tree there are shown locking sessions in nodes below - waiting sessions blocked by sessions in the parent node Details for the selected session Query text Session parameters Sample screen with locks is presented below Locks history Tab [Locks history] tab allows you to analyze locks which have occurred in the past. The screen consists of the following areas: Filter bar by date range Chart showing the locks in time The tree of the blocked session refreshed after clicking in the fragment/the point of the graph 92

93 at the top of the tree - shown locking sessions in nodes below waiting sessions blocked by sessions in the parent node Details for the selected session Query text Session parameters, among others: Transaction opening time, the type of transaction, etc. The sample screen with locks from history is shown below: 93

94 6.9 Parameters Window The system allows you to view and report history of database parameters change over time. The window allows you to view the current status of parameters and their changes over time. Below are examples of screens: State of parameters with control% prefix The history of parameter changes: IMPORTANT: The parameter module is also available from the level of main menu after exiting from the performance module of [Database Analysis]. We go there after clicking [Back to dashboard]. Then system allows to simultaneously analyze parameters for all monitored databases. 94