ST03N Workload Monitor

Workload Monitor PDF download from SAP Help Portal: http://help.sap.com/saphelp_nw70ehp1/helpdata/en/2d/b8be3befaefc75e10000000a114084/frameset.htm Created Creat ed on April 30 , 2014

The documentation may have changed since you downloaded the PDF. You can always find the latest information on SAP Help Portal.

Note This PDF document contains the selected topic and its subtopics (max. 150) in the selected structure. Subtopics from other structures are not included.

© 2014 SAP AG or an SAP affiliate company. All rights reserved. No part of this publication may be reproduced or transmitted in any form or for any purpose without the express permission of SAP AG. The information contained herein may be changed without prior notice. Some software products marketed by SAP AG and its distributors contain proprietary software components of other software vendors. National product specifications may vary. These materials are provided by SAP AG and its affiliated companies ("SAP Group") for informational purposes only, without representation or warranty of any kind, and SAP Group shall not be liable for errors or omissions with respect to the materials. The only warranties for SAP Group products and services are those that are set forth in the express warranty statements accompanying such products and services, if any. Nothing herein should be construed as constituting an additional warranty. SAP and other SAP products and services mentioned herein as well as their respective logos are trademarks or registered trademarks of SAP AG in Germany and other countries. Please see www.sap.com/corporate-en/legal/copyright/index.epx#trademark for additional trademark information and notices.

PUBLIC © 2014 SAP AG or an SAP affiliate company. All rights reserved.

Page 1 of 27

Table Tab le of content cont ent 1 Workload Monitor 1.1 1.1 Operating the Workload Monitor 1.1.1 1.1.1 User Use r Mode Mod e 1.1. 1.1.2 2 Structure of the the Workload Tree in the the Workload Monitor 1.1. 1.1.3 3 Analysis Views of the Workload Monitor 1.1. 1.1.3.1 3.1 Workload Overview Overvi ew 1.1. 1.1.3.2 3.2 Displaying Display ing Transaction Profiles 1.1. 1.1.3.3 3.3 Displaying Display ing Transaction Transactio n Details 1.1. 1.1.3.4 3.4 Displaying Display ing Application Applica tion Statistics 1.1. 1.1.3.5 3.5 Displaying Display ing the Time Time Profile 1.1. 1.1.3.6 3.6 Hitlists: Display 40 Worst Respo Response nse Times Times 1.1. 1.1.3.7 3.7 Displaying Display ing Memory Usage Statistics 1.1. 1.1.3.8 3.8 Displaying Display ing RFC Profiles 1.1. 1.1.3.9 3.9 Displaying Display ing User Profile and Account Statistic Statistics s 1.1. 1.1.3.10 3.10 Displaying Display ing Front End End Statistic Statistics s 1.1. 1.1.3.11 3.11 Displaying Display ing Spool Statistics 1.1. 1.1.3.12 3.12 Displaying Display ing the Respo Response nse Time Distribution 1.1.3.12. 1.1.3.12.1 1 Response Resp onse Times: Rough Guide 1.1. 1.1.3.13 3.13 Displaying Display ing Database Databas e Procedure Proced ure Calls 1.1. 1.1.3.14 3.14 Displaying Display ing Table Access Statistics 1.1. 1.1.3.15 3.15 Displaying Display ing Load from External External Systems 1.1. 1.1.3.16 3.16 Displaying Display ing Web Statistics 1.1. 1.1.3.17 3.17 Displaying Display ing Database Databas e Connection Connection Statistics 1.1. 1.1.3.18 3.18 Displaying Display ing VMC Statistics 1.1. 1.1.4 4 Displaying Display ing a Business Business Transaction Analysis 1.1. 1.1.5 5 Displaying Display ing the Last Last Minutes Load 1.1. 1.1.6 6 Displaying Display ing the Workload History and Workload Distribution in the 1.1. 1.1.7 7 Displaying Display ing Users per Instance Instance 1.2 1.2 Configuring Configuring the Workload Monitor 1.2. 1.2.1 1 Controlling and and Monitoring the the Generation of Statistics Data 1.2. 1.2.2 2 Displaying Display ing and Deleting Contents of the Workload Collector Datab 1.2. 1.2.3 3 Displaying Display ing and Changing Changing the Scheduling of Performance Colle Collectors ctors 1.2. 1.2.4 4 Reorganizing and Monitoring the the Monitoring Databa Database se MONI 1.2. 1.2.5 5 Saving User-Specific Views


Page 2 of 27

1 Workload Monitor Purpose You can use the workload monitor (transaction ST03) to analyze statistical data for the ABAP kernel. If you are analyzing the performance of a system, you should normally start with the workload overview. You can also display the total values for all instances, and compare the performance of particular instances over a period of time. A wide range of analysis views and data helps you to find the source of performance problems quickly and easily. You can use the workload monitor to display the following, among other things: ● Number of instance s configured for your sys tem ● Number of users working on the different instanc es ● Resp onse time distrib ution ● Distribution of workload by transaction steps, transactions, p ackages, subap plications, and applic ations ● Transactions with the largest response times and database time ● Memory usage for each transac tion or each user per dialog step ● Workload caus ed by RFC, broken down by transac tions, function modules, and destinations ● Number and volume of spool requests ● Statistics about response time distrib ution, with or without the GUI time ● Op tional: table access es ● Workload and transactions used b y users, broken down by users, acc ounting numbers, and clients ● Workload generated by reques ts from external systems The following applies to all of this data: ● You can disp lay data for any ins tance (not only the instance that you are logged onto), or optionally summarized for all instances . ● Dep ending on the user mode, you can choose the period for which you want to disp lay data between day, week , and month, or sp ecify the Last Minutes’ Load as you require. ● For most analysis v iews, you can display all or only certain task types .

Features The workload monitor has an interface that is divided into two parts. Use the tree structures on the left to make the following settings: ● Selec t the user mode ● Selec t the period for which y ou want to disp lay the workload ● Selec t different functions and analys is views (whi ch data do you want to disp lay? ) The system then displays the result on the right in a standardized ALV Grid Control. You can use this to: ●

Adjust the layout of the data output

● Find any information using sort and filter functions ● Save user-specific views ● Disp lay statistics graphically More information: Operating the Workload Monitor Configuring the Workload M onitor

1.1 Operating the Workload Monitor Use The Workload Monitor (transaction ST03) is a one-screen transaction that has as few additional menus as possible. This makes operation significantly easier and more intuitive.

Features


Page 3 of 27

Activities The individual screen elements of the workload monitor have the following meanings:

Navigation Area Choose any icon by double-click ing it: Function

Meaning

Depending on the us er mode, you can define the i nstance and the period that you want

Workload

to analyze here. Detailed Analysis

These functions read the workload directly from the statistics files of the individual instances.

Business Transaction Analysis

You can perform a very precise analysis of indivi dual transactions here, down to the level of individual steps.

Last Minutes’ Load

You can u se this function to analyze the u naggregated workload data that has n ot yet been written to the database.

Load History and Distribution

With these fun ctions, you cannot display th e workload for a particular i nstance and a

Workload History

particular period (as in the other workload monitor functions), but rather compare the

Instance Comparison Users per Instance Collector & Perf. Database

workload of different instances or periods. Display th e most im portant data together therefore allows a direct comparison of the instances. Among other things, you can use this function to define which values the statistics collector collects, h ow often, and how l ong they are to be retained in the database in wh at time resolution.

Analysis Views An analysis v iew displays a particular aspect of the workload. In the Analysis View sub screen, c hoose the view that you want to analyze by double-click ing it.

Only the analysis v iews that are active in the selected user mode and for which data exists are disp layed.

Output Area The output area uses an ALV Grid Control, with which you can greatly tailor the selected view to your requirements. For most load parameters, the workload monitor displays more data fields than are required for your analysis. You therefore have a considerable amount of help available in the output area to find the information rele vant for y ou: ● The data of most analysis views i s grouped in the results area by tab pages for different topic areas. ● To obtain an overview of which data fields ex ist for an analys is vi ew, choose the All Data tab page. ● Using the

standard functions of the ALV Grid Control , you can execute the following commands:

○ Show and hide columns ○ Sort rows by the contents of a column ○ Set and delete filters ○ Perform summ ations ○ Export tables as a file type of your choice ○ Display tables as graphics ○ Save sort orders, filters, and selec ted columns as your layout

Editing Views


Page 4 of 27

Choose one of the following buttons in the Edit Views sc reen area: Button

Meaning

Saves the current view as your initial screen for the workload monitor (see also Saving

( Save View )

User-Specific Views) (Previous View)

Moves one view forward or back in the view history

( Next View )

Full Screen On/Off

Shows or hides the navigation bar and the available analysis views on the left of the screen

Workload Monitor start page

1.1.1 User Mode Definition The user mode determines the view of the statistics data of the workload monitor (transaction ST03). The design was conceived with different users in mind, so that the system displays different functions and time periods of monitoring, depending on the user mode you choose.

Use There are two different user m odes. Each s hows different information. · Service Engineer mode This mode provides you with the workload statistics for the current day and for the previous week, as well as an overview of the workload history and distribution, and a detailed analysis of the workload. This user mode is set up especially for SAP Service. By default, the sys tem displays all statistics for all application servers. · Expert mode This mode provides users with all of the functions available in the workload monitor. You can display all available workload data (daily, weekly, and monthly data).

Integration The user mode that you choose determines the functions that are displayed on the initial screen of transaction ST03. Workload Monitor start page

Structure of the Workload Tree in the Workload Monitor Definition In the workload tree, the workload monitor disp lays a hierarchy of two or three levels in the Functions sub tree, depending on the selected user mode. You can use this to quickl y define the instance and peri od for which you want to analyze the workload. You can also optionally use the TOTAL entry (see als o Controlling the Generation and Retention Period of Statistics Data) at instance level. This displays the workload for all instances of the system.

Structure Depending on the selected user mode, you can decide for what period of time and which instance you want to analyze the workload. Mode

Workload Tree Structure

Service Engineer

Previous Week/Today /TOTAL

Expert

/TOTAL

Explanation The system displays a two-level hierarchy in which you select first the period and then the desired instance. The system displays a three-level hierarchy in which you select the instance, then the period to be investigated ( Day , Week , or Month ), and finally the actual time period.


Analysis Views of the Workload Monitor PUBLIC © 2014 SAP AG or an SAP affiliate company. All rights reserved.

Page 5 of 27

Definition An analysis v iew is a particular aspec t of the workload. Depending on which asp ects of the workload you are interested in, you can disp lay various workload parameters, sorted by different sort criteria (such as by users, transaction steps, RFC destinations, and so on).

Use To activate a particular analysis view, choose the desired view in the

Analysis Views subs creen by double-clic king it.

Structure The following analysis views are available: Analysis View

Description

Workload distribution for the individual task types

Workload Overview Transaction Profile

Most im portant performan ce data for all executed transactions; you can also output th e data for sum mari zation levels Packages , Subapplications , or Appli cations Statistics about workload distributi on with in a transaction step (activation at transaction

Application Statistics

level required) Temporal distribution of transaction steps over a day; this allows you to easily identify

Time Profile

periods of higher and lower w orkload Hitlists

The 40 transaction steps with th e highest response tim e and database time; the same hitlists also exist for application statistics Average and maximum memory per dialog step for each transaction and each user

Memory U sage Statistics RFC Profiles

Workload caused by Remote Function Calls on the client- and server-side, broken down by transactions, function modules, or destinations Workload broken down by users, account numbers, and clients

User Profile and Account Statistics

Volum e of data sent and received and the frontend processing times of the i ndivi dual

Frontend Statistics

presentation servers Spool Statistics

Statistics about the nu mber and volu me of spool requests

Response Tie Distribution

Statistics about the distri bution of the response tim es, showi ng what percentage of the system requests were dealt with in what time (with or without GUI time)

Table Access Statistics

Workload, broken down by table accesses (optional)

Load from External Systems

Workload for a system caused by requests from other system s, broken down by systems, actions, or users Statistics about the called database procedures, optionally broken down by the callin g

DB Procedure Calls

transaction Workload caused by Internet calls, on the cli ent- and server-side, broken down by h osts,

Web Statistics

transactions, users, or the URL Workload caused by database operations, broken down by the current database

DB Conn ection Statistics

connection i n each case, or by the calli ng transaction Workload due to the operation of th e

VMC Statistics

VM Container : Resource usage and time usage

by VMC actions (Garbage Collection, and so on)

You can use various analysis v iews, depending on the user mode you have s elected: User Mode Possible Analysis Views

Service Engineer

Workload overview

Expert

Workload overview

Transaction profile

Transaction profile

Application statistics

Application statistics

Time profile

Time profile

Hitlists

Hitlists

Memory u sage statistics


RFC profiles

RFC profiles

User profile Frontend statistics

User profile Account statistics

Response time distribution

Frontend statistics

Load from external systems

Spool statistics

DB procedure calls


Web statistics

Load from external systems

DB connection statistics

DB procedure calls Web statistics DB connection statistics VMC Statistics


Page 6 of 27


1.1.3.1 Workload Overview Definition The workload overview displays data about the workload that is written by the SAP kernel and displayed with the workload monitor (transaction ST03). The workload overview provides system administrators with various detailed information about the most important workload data, such as the CPU time, the number of database changes, the response times, and so on.

Structure You can display the workload overview for all task types ( Dialog, Background, RFC, ALE, and Update ), or only for one particular task typ e. The sys tem displays the following data:

In the case of the HTTP task type, the workload is, by default, not displayed b roken down by app lications or pages (URLs). Instead, SAPM HTTP is used as a dummy name. If you require a relevant breakdown, choose the following path in the navigation bar: Performance DB

→

Workload Collector →

Information

Average CPU time

Collector and

Parameters , and choose the desired radio button in the Created Profiles group box. Explanation

Average time that a work process uses the CPU; during a dialog step, the CPU of the application server is u sed for processing (l oading, generating, processing database requests, ABAP processing, and so on). The CPU ti me i s determined by the operating system. At the end of a transaction step, the SAP work process queries the CPU time from the operating system. The CPU time is therefore not an additive component of the response time, unlike the wait time, roll in time, load time, and database time.

Average response time

Average time between the time at which a dialog process sends a request to a dispatcher work process, and the tim e at which the dial og is complete and the data is transferred to the presentation layer. The response tim e does not include the time for transferring the data from the SAP front end to the application server. Networks w ith bad performance can therefore have larger subjective response times. The transfer time is included in the GUI time and the front end network time.

Average wait time

Average time an unprocessed dialog step waits in the dispatcher queue for a free work process. Under norm al circumstances, the dispatcher work process should pass a dialog step immediately after receiving the request from the dialog step. In this situation, the average wait time should be only a few milliseconds. If the application server or the entire system is under a heavy load, this can lead to traffic jams in the dispatcher wait queue.

Average Load and Generation Time

Average time required to load and generate objects such as ABAP source code and screen i nformation from the database

DB Calls

Number of parsed accesses to the database

DB Access

Number of logical ABAP accesses to data in the database; these accesses are made through th e SAP database interface and are parsed into indivi dual database calls. The ratio of database calls to database accesses is important. If the access to information in a table is buffered in the SAP buffers, no database calls to the database server are required. The ratio of calls/accesses therefore gives a good indication about the efficiency of the table buffering. A good ratio would be 1:10.

GUI Time

Response time between the dispatcher and the GUI during the roundtrips (roundtrips are commun ication steps between the SAP system and the front end durin g a transaction step)

Roll ins

Number of rolled-in user contexts

Roll outs

Number of rolled-out user contexts.

Roll In Time

Processing time for the rolling in of user contexts

Roll Out Time

Processing time for the rolling out of user contexts

Roll Wait Time

Wait time in the roll area. If synchronous RFCs are called, the work process performs a roll out and waits for the end of the RFC in the roll area. The RFC server programs can wait for other RFCs to be sent to them in the roll area.

Average Time per Logical Database Call

Average response time for all commands sent to the database system (in milliseconds). The tim e depends on the CPU capacity of the database server, the n etwork, the buffering, and on the i nput/output capabiliti es of the database server. The access tim es for buffered tables are far faster and are n ot considered here.

You can restrict the display to a particular task type. The task type depends on the task of the associated application process and is an identifying element of every statistics record. There are the following task types:


Page 7 of 27

Task Type

Explanation

AUTOABAP

Automatically-processed report (for example, for monitoring tools)

B.INPUT

Transaction step in batch input mode; it is processed by the dialog work process (update dialogs generated in batch inpu t are always processed synchronously, they belong to the UPDATE task type).

BACKGROUND

Transaction step that was started by a background processing work process.

BUFFER SYNC

A synchronization of the local table buffers regularly requested by the SAP system (the time interval is controlled by the profile parameter rdisp/bufreftime).

DIALOG

Usually a transaction step started online by a user (for example, editor dialogs or manual postings).

RFC

Remote Function Call in the ABAP system; it is processed by the dialog work process

CPIC

Other communication using the CPIC interface; it is processed by the dialog work process

SPOOL

Transaction step of a spool work process

UPDATE

Transaction step of the SAP update task; it is automatically started by the dispatcher on a host with an active update process (update processes are usually install ed on the database host)

UPDATE2

V2 update

ALE

IDoc processing; it is processed by the dialog work process

LCOM

The Fast RFC (fRFC or LCOM-RFC) is a very fast form of data transfer that uses a shared memory pipeline. It is only used in internal communication between ABAP and Java in the SAP Web AS.

HTTP, HTTPS, NNTP, SMTP, FTP

Requests from the ICM that are based on the corresponding Internet protocols

ENQUEUE

Enqueue handler

DIALOG(-)GUI

Dialog without GUI

EXT.PLUGIN

The statistical evaluations of these task types are only relevant for in ternal SAP

AUTOTH ( Auto task handler)

purposes.

RPCTH (Task handler remote procedure call) RFCVMC (RFC inside VMC) DDLOG CLEANUP DEL. THCALL (Delayed task handler call) AUTOJAVA HTTP/JSP, HTTPS/JSP

The task types therefore correspond to the work process types – with the exception of the task types RFC , CPIC , ALE , AUTOABAP , and BUFFER SYNC , which simply represent specific app lications in the dialog work process. TOTAL is the total across all task typ es. Backg round Information The response time is usually split into wait time plus execution time. The SAP response time consists of the following components: Response time = wait time + execution time where: execution time = Generation time during the run + Load time for programs, screens, and graphical user interfaces + Roll times for rolling in the work data + ABAP processing time + Database time + Enqueue time for logical SAP lock processes + CPIC/RFC time + Roll wait time (excluding task types RFC/CPIC/ALE). The CPU time is not an additive component of the response time, but rather the total of the CPU time used by the individual components. The CPU time is therefore additional, separate information about the response time. Workload Monitor start page

1.1.3.2 Displaying Transaction Profiles Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the question of which transactions cause which workload.

Prerequisites If you want to disp lay the transaction details for spec ific transac tions, you need to activate the Transaction Detail Profiles for these transactions first (see: Disp laying Transaction Details).


Page 8 of 27

Procedure 1. Start the Workload Monitor by c alling transaction ST03. 2. In the Workload tree, choose the instance and period to be analyzed. 3. Choose

Analys is Views →

Transaction Profile.

4. Choose the desired analysis view: Standard , EarlyWatch , or Details . 5. The system displays the usual output area, in which the requested transaction profile is disp layed.

Result In the Standard and EarlyWatch analys is views , the output area shows the most important workload data, broken down by transactions exec uted. You can use the Task Type dropdown menu to restrict the view to a particular task typ e. You can choose the different level s of data aggregation using the Aggregation dropdown menu. By default, the system disp lays the Transaction level. Depending on the level that you select, the data is summarized in accordance with the component hierarchy by: · Packages (previously: development class es) · Subapp lications (BC-CCM) · App lications (BC, FI, CO, and so on) You can display more information for a data record by double-clicking the line for the corresponding transaction profile. The system opens a second window to display this detailed data. You can also choose different levels of data aggregation here – although the level in the detailed view can be at most one level beneath the level in the main view. The entries that you cannot choose in the dropdown menus are grayed out.

You require a special authorization (profile S_TOOLS_EX) to display user-specific detailed data. As long as you are disp laying daily data for the current day, the system disp lays the Single Records button. You can use this button to directly call the transaction STAD with the selection parameters Instance , Transaction/Report , and, if necessary, User . There are three transaction profiles with the following special features: Transaction Profile

Special Features

Standard

None

EarlyWatch

· Different combination of the individual v alues · No task type selec tion · You can switch between average values per transaction step and average values Logical Unit of Work (LOW) of a transaction

Details

Analysis of individual transaction steps of a transaction (seeDisplaying Transaction Details)


1.1.3.3 Displaying Transaction Details Use The dialog step statistics in the workload monitor allow you to analyze individual transaction steps of a transaction by their function codes. This analysis does, however, require detailed k nowledge of the function codes of the relevant ap pli cation. Therefore, it is only makes sens e to activate the Transaction Detail Profile when analyzing actual p roblems and it is deactivated by default. There are various parameters for the dialog step statistics that you can set for specific analyses if required. Note, however, that activate these parameter means that significantly more statistics records are written and that the load caused by the statistics collector can significantly increase.

You should therefore change the values of the parameters only temporarily for individual analyses, since the number of statistics records can significantly increase, and this c an lead to performance problems in the statistics collector.

Prerequisites To display dialog step statistics for individual transactions, you first need to activate this: 1. Start the Workload Monitor by c alling transaction ST03. 2. Choose the Expert user mode. 3. Choose

Collector & Performance DB →


Parameters

4. The Selec tion of Data to Cumulate dialog screen app ears. Enter the transac tion codes for the transactions to be analy zed in detail in the Create transac tion detail profiles for group b ox. 5. Sav e your changes .

Procedure 1. Start the Workload Monitor by c alling transaction ST03. 2. Choose the instanc e and period to be analyzed by double-c lic king.


Page 9 of 27

3. Choose

Analysis Views →

Transaction Profile →

Details →

.

4. The system displays the usual output area, in which the requested workload is disp layed.

If no dialog step statistics are available for the specified instance or period, the system does not display the icon for the details. Workload Monitor start page

1.1.3.4 Displaying Application Statistics Use The application statistics allow you to analyze resource consumption in more detail than the workload monitor . Using spec ial calls within the ABAP c oding, the sys tem collects statistics for individual parts of an application. While the workload statistics always analyze a complete dialog step, you can use the application statistics to analyze the resources used b y an individual function within a dialog step (such as price determination). In some cases, the app lication statistics c an therefore replace an SQL trace.

The statistical data collected by the application statistics is the same as the workload statistics (response time, and so on).

Prerequisites To use the application statistics, corresponding function calls must be implemented in the application programs. No data can be collected for application statistics without these calls. You also need to ensure that the following profile parameter values are set: stat/as_level = 1 stat/as_collect = 2 You must also have scheduled the collector for the application statistics. In addition, you need to activate the individual components of the application statistics, by calling transaction ST03 and choosing


Statistics Records + File ®

App lication Statistics .

Procedure 1. Start the Workload Monitor by c alling transaction ST03. 2. In the Workload tree, choose the instanc e and period to be analyzed. 3. Choose

Analysis Views →

App lication Statistics.

4. The system displays the usual output are, in which the requested workload is displayed.

If no application statistics are available for the specified instance or period, the system does not display the icon for application statistics.

Result The output area contains the desired workload data, broken down by transaction. To display the data broken down by users, double-click a line.

You require a spec ial authorization (profile S_ TOOLS_EX) to display user-specific data. Workload Monitor start page

Displaying the Time Profile Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the question of how the workload is distributed over the hours of the day. The time profile displays the temporal distribution of transaction steps over a day, meaning that you can easily identify periods of higher and lower workload. The time profile therefore displays the workload grouped by the hours of the day.


Analysis Views →

Time Profile .

4. The system display s the usual output area, in which the requested distribution of the response times is dis played in an


ALV Grid Control.

Page 10 of 27

Result The output area displays the most important data about the workload, broken down by the hours of the day. By default, the Time Profile analysis v iew displays all task typ es. By selecting an entry in the Task Type dropdown menu, you can restrict the display to the desired task type. You can display the executed transactions in the relevant period by double-click ing a line. Workload Monitor start page

Hitlists: Display 40 Worst Response Times Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the question of which dialog steps and application steps have the longest response and database times. These are the hitlists. There are two hitlists of this type for the dialog statistics and for the application statistics: · Hitlists for Dialog Step Statistics These hitlists list 40 transaction steps for each task type. There are hitlists for the transaction steps with the largest response time and those with the largest database time. · Hitlists for Application Statistics These hitlists lis t 40 ap plic ation statistics s teps for each task typ e. There are hitlists for the applic ation statistics s teps with the largest response time and those with the largest database time. The hitlists therefore allow you to identify the transaction steps that use most resources. Using the hitlists, you can answer the following questions, among others: · Is it always the same program that has high resp onse time? · Is this program running at the same time in sev eral sessions (locking work processes)? · How many k iloby tes of information were transferred? · Could the program be started at night in the bac kground?

Procedure 1. Start the Workload Monitor by c alling transaction ST03. 2.

Choose the app ropriate user mode. You can essentially display hitlists in any user mode, but the time period is also restricted for hitlists through the selection of the user mode.

3.

In the Workload tree, choose the instance and period to be analyzed.

4. Choose 5.

Analysis Views →

Hitlists .

Choose the desired hitlist.

6. To display more details, choose a line in the table b y double-clicking it. Workload Monitor start page

1.1.3.7 Displaying Memory Usage Statistics Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these perspectives is the question of the average or maximum values for the memory usage for a particular transaction or a particular user per dialog step.

Procedure 1. Start the Workload Monitor by c alling transaction ST03. 2. Choose the app ropriate user mode. Essentially, you can display the memory usage statistics in any user mode, but the restriction of the time period due to the selection of the user mode also applies to the memory usage statistics. 3. In the Workload tree, choose the instance and period to be analyzed. 4. Choose

Analysis Views →

Memory Use Statistics.


Result The output area contains two tab pages with the following characteristics:


Page 11 of 27

Tab Page

Properties

Transactions

The memory usage is sorted by the transactions executed during the analysis period. Double-click a line to display the m emory u sage of the selected transaction sorted by user.

User

The memory usage is sorted by the user logged on during the analysis period. Doubleclick a line to display th e mem ory usage of the selected user, sorted by transactions.

A sp ecial authorization (profile S_TOO LS_EXE) is required to display user-specific data. Workload Monitor start page

1.1.3.8 Displaying RFC Profiles Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the workload caused b y Remote Function Calls (RFC). By displaying the RFC profiles, you can answer the following questions: · Which transac tion, which function module, and which us er caus ed what workload through RFC calls as an RFC clien t or an RFC server? · What workload do the transactions, function modules, or users c ause in whic h local or remote RFC destinations? For transaction steps with RFC, the kernel writes subrecords with additional information about the processing time, the destination, and the function module used. The parameter stat/rfcrec (default: stat/rfcrec = 5) sp ecifies the maximum number of subrecords of each type (RFC client, RFC server, RFC client destination, and RFC server destination) that the kernel writes. If more RFCs are performed during a transaction step, only the five calls with the longest execution time are therefore logged, Which is sufficient for performance analyses. The restriction to a maximum of five records represents a compromise between the required accuracy on one side and the workload created by the performance collector on the other side. A larger value for stat/rfcrec can lead to performance problems i n the collector (see also Configuring the Workload Monitor ). The RFC client and RFC server records contain data such as execution time and called function for individual RFCs. The RFC destination records contain the total of all RFC calls per destination and therefore no additional information about the called function modules.


Analysis Views →

RFC Profile , and choose the desired RFC profile.


Result Depending on the RFC profile selected, the output area shows the following tab pages: Tab Page

Description

· Function Module · Transaction · User

What workload is caused by the function modules, transactions, or users (depending on

· Remote Des tination · Remote Server · Local Server

Where is th e RFC w orkload created?

the selected RFC profile, as the RFC client or the RFC server)?

A sp ecial authorization (profile S_TOO LS_EXE) is required to display user-specific data. On all tab pages: if you double-click a line, a new screen appears showing detailed information about the selected line. Workload Monitor start page

Displaying User Profile and Account Statistics Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the question of how the system load is distributed among the different users and what the workload profile of an individual user looks like. With this information, you can answer questions suc h as: · How long is the resp onse time that a user must wait on average how is this time s pli t? · How many dialog steps has a particular user executed?


Page 12 of 27

· Which database accesses with what workload did a particular user cause?

You require a sp ecial authorization (profile S_TOOLS_ EX) to display this analysis view.

Procedure 1. Start the Workload Monitor by c alling transaction ST03. 2. Ensure that you are in the Expert user mode. 3. In the Workload tree, choose the instance and period to be analyzed. 4. Choose

Analysis Views →

User and Settlement Statistics , and choose the desired analysis view.


Result User Profile Analysis View The output area contains five tab pages that display the most important workload data for individual users. To see a breakdown of the workload caused by a user, sorted by the individual transactions, double-click the user. If you have chosen TOTAL for the Workload , the sys tem disp lays the workload data for indivi dual users totaled for all instanc es. To see a break down of the workload caused by a user, sorted by the individual instances, double-click the user.

Settlement Statistics Analysis View The output area displays the corresponding workload in this sy stem on the tab pages Clients , Account Numbers , and Account Numbers . Every user is usually assigned to an account number in the user master record (see also:

User Maintenance Functions). To view information about the individual users of a

client or account number, double-click the client or account number. Workload Monitor start page

1.1.3.10 Displaying Front End Statistics Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the question of how the workload is distributed among the various presentation servers and instances. With this information, you can answer questions such as: · What volume of data is exchanged between the individual pres entation servers and SAP ins tances? · Which res ponse times arise at the front end (GUI, front end network, and s o on)? · Which pres entation servers are causing what workload in a partic ular instance?


Analys is Views →

Frontend Statistics .


Result The output area contains two tab pages with the following characteristics: Tab Page Frontends

Description The workload is displayed, sorted by the presentation servers of the selected instance. The values displayed are for the aspects of the workload that arise in connection wi th the presentation server, such as the number of communication steps, transferred data value, or front end wait times.

Instance(s)

If you have chosen the TOTAL workload (see alsoControlling and Monitoring the Generation of Statistics Data), the sam e aspects of the workl oad are displayed, sorted by instances of the system. To display the workl oad for the presentation servers assigned to an in stance, double-click that i nstance.



Page 13 of 27

1.1.3.11 Displaying Spool Statistics Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the question of how much workload occurs due to spool requests. With this information, you can answer questions such as: · How many sp ool requests with which properties were ex ecuted on the selected instance or the entire sys tem? · How long was the average resp onse time and the average wait time for spool requests? · Which and how many logical transac tion steps were there in spool work processes ? The spool statistics are therefore useful for spool administrators who want to check what was output in each application server.

The analysis view Spool Statistics may not be availab le for every configured application server in your SAP system.

Procedure 1. Start the Workload Monitor by c alling transaction ST03. 2. Ensure that you are in the Expert User Mode. 3. In the Workload tree, choose the instance and period to be analyzed. 4. Choose

Analysis Views →

Spool Statistics.

If no spool statistics are available for certain instance, the system does not display the Spool Statistics icon.

Result The output area contains two tab pages with the following characteristics: Tab Page

Meaning

Overview

The most important workload data for the executed spool requests

Spool Activity

The workload of the spool work processes


Displaying the Response Time Distribution Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the question of how response times are distributed with respect to the different task types. The response time for the transaction steps is split into 10 categories: t < 0.1 s; 0.1 s < t < 0.2 s, and so on You can choose between two different types of display: · Number of transac tion steps (abs olute or perc entage) that are in a particular response time category (such as a resp onse time between 0.4 s and 0.5 s ). · Percentage of transaction steps whose response time is not longer than the upper limit of a particular response time category (such as a response time between 0 s and 0.5 s). This analysis view therefore allows you to check quickly what percentage of all transaction steps have a response time below a certain threshold value. For more information about what response time you can expect, see: Response Times: Rough Guide.

Since the response time has included the communication between the SAP s ystem and the presentation server s ince SAP R/3 4.6 (see also SAP Note 2039 24), you can display the response time for the Dialog task type either including or excluding the GUI time in this analysis v iew. In this way, you can remove the influence of the performance of the front end systems and the network from the calculation of the response time distribution.


Analysis Views →

Response Time Distribution .

4. The sy stem disp lays the usual output area, in which the requested distrib ution of the workload is dis play ed.

Result PUBLIC © 2014 SAP AG or an SAP affiliate company. All rights reserved.

Page 14 of 27

The output area contains three tab pages with the following characteristics: Tab Page

Description

Steps

The system displays the absolute number of transaction steps in the individual response time categories, broken down by task type.

Percentage

The system displays the proportion of transaction steps (as a percentage) in the indivi dual response tim e categories, broken down by task type.

Percentage (Cumulated)

The system displays the proportion of transaction steps (as a percentage) with a response time between 0s and the upper limit of the individual response time categories, broken down by task type.


1.1.3.12.1 Response Times: Rough Guide Definition The following times provide you with an overview of the optimal response times in your SAP system. You can display the current workload overview for your system with the workload monitor. Proportions of the Response Time Performance Data

Time

Average response time

Around 1 second (dialog), < 1 second (update)

Average CPU time

Around 40% of the average response time

Average wait time

< 1% of the average response time

Average l oad time

< 10% of the average response tim e

Average database time per dialog step

Around 40% of the average response time

- The operating sys tem can affect these values by up to around 10%. - The ratio of the CPU time to the total time should not fall under 5% . A sm aller p roportion could mean that there is an input/output b ottleneck or that a database overload has caused long wait times. - The average response times for dialog transactions (online transactions) should not be more than a second, sinc e a higher value would strain the patience of the user too much. Times for Database Access Database Access

Time

Direct read

< 10 ms

Sequential read

< 40 ms

Changes

> 25 ms

Possible Causes of Bad Response Times High Value For

Possible Causes

DB request (Change/Comm.)

Problems wi th database or index

Load time

· Buffer too small · Transports to the productive sys tem

Wait time

· · · · ·

Insufficient work proces ses Inactiv e update Incorrect operation mode Locked tasks Long-running transactions

- In a databas e sys tem that is functioning properly , the times for direct reads and changes should not be more than 10 mill isec onds. - The time for sequential reads s hould not be more than 30-40 milli sec onds. Workload Monitor start page

Displaying Database Procedure Calls PUBLIC © 2014 SAP AG or an SAP affiliate company. All rights reserved.

Page 15 of 27

Use In the workload monitor, you can display the workload in different analysis views that highlight the workload from a particular perspective. One of these aspects is the question of how much load is generated due to calls of

database procedures.


Analysis Views →

DB Procedure Calls .



Meaning

DB Procedure

Workload due to database procedure calls, broken down by the different database procedures. To display the workl oad broken down by different transactions or reports that called a specific procedure, select that procedure.

Transactions

Workload due to database procedure calls, broken down by the different transactions or reports. To display the w orkload broken down by different database procedures that were called by a specific transaction, select that transaction.


1.1.3.14 Displaying Table Access Statistics Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the question of which transactions have accessed which database tables, and how often. Since the number of statistics records greatly increases by logging table acces ses of this typ e, this option is deactivated by default. You should only ac tivate it temporarily for individual analyses, and for individual transactions, since performance problems can occur in the statistics collector otherwise.

Prerequisites To display table access statistics for individual transactions, you first need to activate this: 1. Start the Workload Monitor by c alling transaction ST03. 2. Ensure that you are in the Expert User Mode. 3. Choose


Statistics Records & File →

Online Parameters →

Dialog Step Statistics .

4. The Change Runtime Parameters of Statistics dialog screen appears. You can change the online parameter stat/tabrec , which sp ecifies the maximum number of table acc ess statistics subrecords that the kernel is to write for each transaction step. The N most resource-intensive accesses of a transaction step are stored. If you only want to activate the table access statistics for certain transactions, you can specify a maximum of five transactions in the online parameters stat/tcode1 to s tat/tcode5 . The table acces s statistics are then only produced for these transactions.

Change the parameter stat/tabrec only temporarily for indiv idual analys es. It must be set to 0 (zero) at all other times. Otherwise, the numb er of statistics records increases s ignificantly, which can lead to performance problems i n the statistics collector.


Analysis Views →

Table Acces s Statistics.

If no table acc ess statistics are available for certain instance, the sys tem does not display the icon for the table access statistics. 4. The sys tem displays the usual output area, in which the requested table acces s statistics are displayed.

Result The output area contains two tab pages with the following characteristics:


Page 16 of 27

Tab Page

Description

Tables

The database tables that the activated transactions accessed are displayed with the most im portant load data (nu mber of selected, read, and m odified data records, access tim e). If you choose a table by double-clicking it, another screen appears, sh owing th e transactions that accessed this specific table.

Transactions

The activated transactions are displayed with the most important load data concerning table accesses (see above). If you choose a transaction by double-clicking it, another screen appears, showin g the tables that this specific transaction accessed.


Displaying Load from External Systems Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the question of how much workload in a system arises on the basis of actions from other systems.

If there is no connection to another SAP system in a system landscape, this analysis view does not exist.


Analys is Views →

Load from External Sys tems.

4. The system displays the usual output area, in which the requested workload statistics are displayed.

Result The output area displays the most important workload data that was triggered by actions in external systems, grouped by different points of view on five tab pages. Each line c ontains the data from a s pecific external system. You can break down the list further by action and user us ing the Aggregation dropdown menu. You can dis play additional details b y double-click ing a line.

A sp ecial authorization (profile S_TOO LS_EXE) is required to display user-specific data.

Background Information When a transaction is started, a unique ID, the trans ID, is generated. This is transferred together with the data during Remote Function Calls (as of SAP R/3 4.6D). This ID is retained even for RFCs over multiple systems, meaning that it is possible to identify the system that is at the start of the chain and that originally triggered the action. Workload Monitor start page

1.1.3.16 Displaying Web Statistics Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the question of how much workload occurs due to requests from the Internet. With this information, you can answer questions such as: · What quantities of data are sent to which URL, how often, and how much time is requi red to do so? · Which host/p ort combi nations function as Web serv ers, and what workload is generated?


Analysis Views →

Web Statistics.

4. Choose the desired statistics : Statistics

Description


Page 17 of 27

WEB Client Statistics

Workload due to requests for which the system acts as a Web client

WEB Client Dest. Statistics

Workload due to requests for which the system acts as a Web client, broken down by different client destinations

WEB Server Statistics

Workload due to requests for which the system acts as a Web server

WEB Server Dest. Statistics

Workload due to requests for which the system acts as a Web server, broken down by different server destinations


Result The output area contains four tab pages with the following characteristics: Tab Page

Description

Host

Breakdown of the workload by host names and port numbers that act as Web servers or Web clients

Transactions

Breakdown of the workload by reports or transactions that act as Web servers or Web clients

User

Breakdown of the workload by users and clients that act as Web servers or Web clients

URL

Breakdown of the workload by URLs that act as Web servers or Web clients


Displaying Database Connection Statistics Use In the workload monitor, you can display the workload in different analysis views which highlight the workload from a particular perspective. One of these aspects is the question of how the workload generated by accesses to databases is distributed across the various database connections.


Analysis Views →

DB Connection Statistics .



Connections

Meaning

Workload due to accesses to the database, broken down by different database connections. To display the workload broken down by different transactions or reports that used a specific database connection, select that database connection.

Transactions

Workload due to accesses to the database, broken down by different transactions or reports. To display the w orkload broken down by different database connections u sed by a transaction, select that transaction.


1.1.3.18 Displaying VMC Statistics Use In the workload monitor, you can display the workload in different

Analysis Views, which highlight the workload from a particular perspective. One of these

views shows how the availab le resources are made us e of at spec ific times when the VM Container is operating. With this information, you can answer questions such as: How often has the VM Container been called up? How often has the garbage collector been called for the various memory areas (old generation, young generation, shared objec ts)? (see Memory


Page 18 of 27

Management in the VM Container ) How much time has the system spent in the various garbage collections? How many IT compilations have there been?

Procedure Start the Workload Monitor by calling transaction ST03. 1. Ensure that you are in the Experts

User Mode.

2. In the Workload tree, choose the instance and time period to be analyzed. 3. Choose Analysis Views → VMC Statistics . The system disp lays the usual output area, in which the requested workload is disp layed.

Result The output area shows the workload in this system in the corresponding tab pages, Times , Memory , Shares, Parts of Response Time and All Data . You can get information about these display modes via the F1 help. You can find information about the memory types, garbage collections, and so on, in Memory Management in the VM Container .

Displaying a Business Transaction Analysis Use You can use the workload monitor to display the business transaction analysis. The analysis calculates the system resource usage of individual transactions and returns the most detailed analysis of a transaction and the dialog steps. The selection criteria include user, transaction, program, task type, start date, and start time. The business transaction analysis uses the statistics files STAT and ASTAT (dialog step statistics and app lication statistics) in the file sy stem of the relevant instances as its data sources. This data basis allows more detailed analysis than the other analysis views of the workload monitors, which use the aggregated data from the workload collector database as their data source. Note that you can only execute the business transaction analysis for the period for which the statistics files are retained (by default, 48 hours).

The analyzed time period can be larger than the interval defined by Read Time , as the system analyzes complete transactions as far as possib le. However, it is not always possib le to perform a c omplete analysis for long-running transactions. As the b usiness transaction analysis is time-consuming, you should use as short an interval as possib le (around 10 minutes).

Procedure 1. Call transaction ST03 to start the Workload Monitor. 2. Choose Detailed Analysis → B usiness Transaction Analysis . 3. The Select Statistical Records dialog screen appears . You can make the following settings here: Choose one of the following disp lay modes in the Dis play M ode group box: ¡ Show all statistics records, sorted by start time The system displays a chronological list of dialog steps with their resource usage (this corresponds to the individual statistics records and the transaction statistics). ¡ Show all records, grouped by b usiness transaction The system displays a list of dialog steps with their resource usage, grouped by transactions. The list contains the individual dialog steps of a transaction. ¡ Show business transaction sums The system displays a list with totals of workload data for every transaction. This variant is used most frequently.

In the first two cases – if you have restricted the data to certain users or transactions, and so on – the system reads only the corresponding records. In the third case, the system reads all records, but displays only the records that match the selection criteria. You should consider this if you switch to a different display mode by choosing Disp . Mode . Use filters in the Filter Parameter group box to define which statistics data you want to disp lay. In the Tools group box, c hoose one of the following commands: ¡ Include statistic s from memory The statistics data is buffered first, before it is read into the STAD file. This option forces the buffer to write the data to a file before the analysis is performed. This ensures that you can analyze the most recent transactions. ¡ Include applic ation statistics Choose this check b ox to display appli cation s tatistics in addition. ¡ Server selec tion This pushb utton restricts the analysis to certain instances in the system. Select the desired instances. ¡ Additional options You can set the wait time for the Remote Function Call (RFC), maximum runtime of the analysis, and the time period to be used for the analysis. Workload Monitor start page


Page 19 of 27

Displaying the Last Minutes’ Load Use The workload monitor usually reads the workload data from the workload collector database, which is supplied with data from the statistics files for the individual instances once an hour. This means that you usually cannot show any data that is less than an hour old and has not yet been written to the database. So that you can also monitor the latest changes in the sys tem performance, there is an additional function, Last Minutes’ Load . You can us e this function to directly acces s the statistics files for the individual instances and to display the system p erformance for any p eriod (such as for the last 15 minutes). After you have selected an instance, the usual analysis views are available with data for the selected period.

Procedure 1. First start the Workload Monitor by call ing transaction ST03 . 2. Expand the Last Minutes’ Load sub tree and double-c lic k the instanc e for which you want to inves tigate the sy stem performance. 3. The Last Minutes Load on dialog screen app ears. You can make more precis e spec ifications about the workload that you want to analyze on this screen. 4. The system displays the usual output area, in which the requested workload is disp layed. Workload Monitor start page

Displaying the Workload History and Workload Distribution in the Workload Monitor Purpose The workload monitor is conceived in such a way that the workload for a particular instance that you can display the workload for a specific instance and a particular period of time, and compare the load for these circumstances caused by various transactions, users, RFC destinations, external systems, and so on. Some questions cannot, however, be answered with these disp lay conditions, suc h as the development of the workload over time ( Load History ) or the distrib ution of the workload for different instances ( Instance Comparison ).

Process Flow Workload History Use the Load History to check how the workload develops for a selected task type and a selected instance. 1. Start the Workload Monitor by c alling transaction ST03. 2. Expand

Load History and Distribution →

Load History .

3. The system displays a lis t of the available instances and optionally the TOTAL entry for all instances (see also Controlling the Creation and Retention Period of Statistics Data). Choose the desired instance b y double-click ing it. 4. The Workload: Compare all The Workload: Compare all peri ods for one instanc e sc reen appears . The sys tem displ ays a list of workloads for the selec ted instance at different times. You have the following options on this screen: ¡ Changing the Aggr egates You can disp lay the workload for different Day s , Weeks , or Months for the selec ted instance. Choose the corresp onding button to do this. ¡ Changing the Output View You can use the output view to select which workload data is displayed on the screen. You have the following options: Output View

Mean Values Total Values

Description

Selected workload values from theWorkload Overview analysis view, for the selected task type in each case

DB Access Response Time Distribution

See Displaying the Response Time Distribution

Measurement Interval

Time stamp of the first and last statistics records during the selected period

¡ Sorting the Output You can sort the output by any c olumn. To do this, s elect the desired column and choose the Sort button. ¡ Changing the Task Type You can speci fy the task type for which workload is to be disp layed. To do this, choose Other Task Type , and select the desired type.

Instance Comparison In the Load History , you displ ayed the workload data for an instance for different times; in the Instance Comparis on , you disp lay the workload data for different


Page 20 of 27

instances for one period. 1. Start the Workload Monitor by c alling transaction ST03. 2. Expand

Load History and Distribution and then double-click

Instance Comparison .

3. The Workload: Compare all The Workload: Compare all instanc es for one period sc reen appears . The sy stem displ ays a list of all instances and their workloads during a particular period. You can use the following commands: ¡ Changing the Output View You can use the output view to selec t which workload data is displ ayed on the screen. The same output views are avai labl e as in the Load History . ¡ Sorting the Output You can sort the output by any c olumn. To do this, s elect the desired column and choose the Sort button. ¡ Changing the Task Type You can speci fy the task type for which workload is to be disp layed. To do this, choose Other Task Type , and select the desired type. ¡ Changing the Period You can determine the time period for which workload data is to be displ ayed. To do this, c hoose the Previ ous Period and Next Period buttons (you can select a period of Days, Weeks, and Months by choosing Edit → Choose Period Type ). The Load History and Instance Compari son disp lay only data that you can selec t using the Analysis Views. The difference is that in the Load History, you can compare workloads directly in one list, which would otherwise always be in different output tables due to the structure of the workload monitor. It is therefore a summary – the system displays the same data in a different format.

Users per Instance Another comparison that you can display for workload distribution is the number of users p er instance. Workload Monitor start page

1.1.7 Displaying Users per Instance Use You can use the workload monitor to display the number of users working on an instance. You can display the number of dialog steps that each user has executed, and check that the instance response times are accep table.

Procedure 1. Start the Workload Monitor by c alling transaction ST03. 2. Choose

Load History and Distribution →

Users per Instance.

The Performance: Display instances and their users screen app ears. This s hows a lis t with the number of users logged on to each instance. You can use the following commands to obtain more information: ¡ Instances/User and Users per Instance Button to switch between two views (the button always shows the view to which you can switch, not the active view): Button

Meaning

Instances/User

For each user, the system shows all instances that the user logged on to during the selected tim e period.

Users per Instance

For each instance, the system shows the number of users that were logged on during the selected time period.

¡ Switching the Time Period You can determine the time period for which workload data is to be displ ayed. To do this, c hoose the Previ ous Period and Next Period buttons (you can select a period of days, weeks, or months by choosing the By Days , By Weeks , or By Months buttons). ¡ Displaying the Users on an Instance You can displ ay the users on a particular instance by double-clicking the desired instance in the Users per Instance view. The system display s another list of the users on that instance. To then disp lay the corresp onding data for another instanc e, choose the Previ ous Instance or Next Instance buttons. Workload Monitor start page

1.2 Configuring the Workload Monitor Purpose You use the workload monitor to analyze statistical data from the SAP kernel. You can influence the type and scope of the data entry and the data storage in a number of ways. This process presents these influence methods.

You do not need to configure the Workload Monitor before you use it for the first time. The delivered settings usually ensure correct operation. You therefore only need to change the parameters of the collector and of the performance database when troubleshooting or if performance problems occur.


Page 21 of 27

Prerequisites Note that to display statistical data, you need to have scheduled the job SAP_CO LLECTOR_FOR_PERFMONITOR (see

Background Jobs to Be Scheduled

Regularly).

Process Flow You can intervene at various points in the data collection and storage:

Controlling and Monitoring the Generation of Statistics Data You can determine whether statistics data is to be saved in the workload collector database at all, broken down by different profiles (which correspond to the various analysis views). You can also determine in the same way how and after how long this data is to be reorganized. For more information, see Controlling and Monitoring the Creation of Statistics Data.

In addition to the workload collector database, there is also the monitoring database MONI, which contains performance data that is displayed in other transactions. You can, however, monitor the scope and reorganization of MONI from the workload monitor (see Reorganizing and Monitoring the Monitoring Database MONI).

Displaying and Deleting Contents of the Workload Collector Database If problems occur when reorganizing the statistics data, you can display and manually delete the corresponding contents of the workload collector database (see Displaying and Deleting the Contents of the Workload Collector Database ).

Displaying and Changing the Scheduling of Performance Collectors Various performance collectors are responsible both for the collection of the workload data and for collecting performance data for other transactions. The collectors run at regular intervals. These times are saved in the table TCOLL. By changing the entries in TCOLL, you can determine which collectors run, and how often (see Displaying and Changing the Scheduling of Performance Collectors).

Displaying and Changing Profile Parameters for Statistics Records There are a number of relevant profile and online parameters that are relevant for the workload collector. You can display the relevant profile parameters separately for each instance by choosing


Statistics Records and File →

Relevant Profile Parameters in the

navigation bar. Since you cannot change these profile parameters during running operation, but only by a restart of the relevant instance, you can only display the parameters . To change the parameters, us e transaction RZ10 ( Edit Profiles ). The two following profile parameters are esp ecial ly imp ortant: ● stat/max_files spec ifies how many statistics files the kernel uses. If the value of the parameter is greater than 0, a new statistics file is c reated for each hour. The size of the value determines how many statistics files are to exist simultaneously before the oldest file is deleted. The default value of the parameter is 48. This means that you can access the statistics data of the dialog step statistics within two days. ● stat/as_max_files spec ifies the corresponding value for the applic ation statistics. Online parameters are profile parameters for which changes take effect immediately (without a restart of the affected instance). Unlike the relevant profile parameters , you can change the online parameters directly in the workload monitor. To do this, choose & File

→


Statistics Records

Onli ne Parameters . You require the online parameters for the following statistic s, among others:

● Table Access Statistics ● App lication Statistics Workload Monitor start page


Page 22 of 27

Controlling and Monitoring the Generation of Statistics Data Use At every transaction step, ABAP work processes first write statistics and app lication statistics data to a buffer. This data is then saved in file system files on the host of the relevant instance. Separate files of this type are created for the statistics and the application statistics every hour. The total number of files is usually restricted to 48 (profile parameters stat/max _files or stat/as_ max_ files ). This data is therefore stored for two days b efore it is overwritten. At the next level, this statistics data is aggregated and saved as daily aggregates (that is, with the time resolution of days) in the workload collector database. A corresponding report runs once an hour on every instance for this. You can define yourself the profiles for which statistics data is to be saved in the database. These profiles each cover a separate topic area and represent the existing Analysis Views of the Workload Monitor . In detail: Profile

AA

Application profile

AB

Application hitlist: DB accesses

AC

Application hitlist: response times

EA

Component hierarchy

Analysis View

Application Statistics Hitlists → Top DB Accesses functions Hitlists

→ Top Response Time functions

This data does not create a separate analysis view, but rather provides additional information for other analysis

EB

Orgunit

WA

Task types profile

Workload overview

WB

Time-dependent task types profile


WC

Time, program, task types profile

Time profile

WD

Database procedures profile

WF

External Systems

Workload from External Systems

WG

Function code profile

Transaction details

WH

Frontend profile

Frontend statistics

WJ

Memory usage profile


WK

Spool activity profile

Spool statistics

WL

Table profile

Table access statistics

WM

User, transaction profile

Transaction profiles and User profile and accountin g

views.

DB procedure calls

statistics → user profile 'WN'

User profile

User profile an d accountin g statistics → Accounting statistics

WO

RFC client profile

RFC profiles → RFC client profile

WP

RFC client destination profile

RFC profiles → RFC client destination profile

WQ

RFC server profile

RFC profiles → RFC client profile

WR

RFC server destination profile

RFC profiles → RFC server destination profile

WT

Hitlist: database accesses

Hitlists → Top DB accesses dialog steps

WU

Hitlist: response times

Hitlists

WV

Spool profile

Spool statistics ( Overview tab page)

WW

HTTP client profile

Web statistics

→ WEB client statistics

WX

HTTP client destination profile

Web statistics

→ WEB client destination statistics

WY

Database connection

DB connection statistics

→ Top response time dialog steps

The statistics data at file system level contains the statistics records as they are written by the kernel after each transaction step, while the data in the workload collector database is already aggregated, meaning, for example, that it is no longer possible to analyze parts of a transaction. This means that after two days, you can no longer perform an analysis below transaction level. The daily aggregates are also further aggregated, in two different ways: · The statistics data of all instanc es in a sys tem is combined to a TOTAL instance, whic h you can displ ay as an additional instance of a sys tem in the workload monitor. · The daily aggregates are combined into weekly and monthly aggregates. You can also define in this case the profiles for which the weekly and monthly aggregates and the aggregates for the TOTAL instance are to be created.

By default, daily, weekly, monthly, and TOTAL aggregates are generated for all profiles. You can define the retention periods for all of these aggregates, broken down by profiles.

Activities To change the generation and retention period of the statistics data, or to monitor the correct functioning of the corresponding collectors, start the workload monitor (transaction ST03) and ensure that you are in the Expert user mode.


Page 23 of 27

Creating the Daily Aggr egates · To determine the profiles for which daily aggregates are to be created in the workload collector database, c hoose the following path in the nav igation bar: Collector and Performance DB →


Instance Collector →

Control . You can only display data in the workload monitor

for profiles that are activated in this table. · To chec k that the corresp onding collector is functioning correctly, choose the following path in the navigation bar: Workload Collector →

Instance Collector →

Collector and Performance DB →

Log . You can view the corresp onding error logs, broken down by time periods. The collector runs

every hour on every instance of the system. Creating the Weekly, Monthly, and TOTAL Aggregates · To determine the profiles for which the above aggregates are to be created in the workload collector database, choose the following path in the navigation bar: Collector and Performance DB →


TOTAL Collector →

Control .

· To chec k that the corresp onding collector is functioning correctly, choose the following path in the navigation bar: Workload Collector →

TOTAL Collector →


Log . You can view the corresp onding error logs, broken down by time periods. The collector runs

system-wide once every three hours. Retention Periods of the Different Aggregates · To determine how long the different aggregates are to be stored in the workload collector database, choose the following path in the navigation bar: and Performance DB →

Performance Database →

Workload Collector Database →

Reorganization →

Collector

Control .

Note that, depending on the size of the screen, the important part of the column headers may not be visible, and the time unit for the different aggregates is different. The table consists of the following columns: No.

Heading

Unit

1

Daily Aggregates Retention Period

Day

2

Weekly Aggregates Retention Period

Week

3

Monthly Aggregates Retention Period

Month

4

Daily Aggregates Retention Period (TOTAL)

Day

5

Weekly Aggregates Retention Period (TOTAL)

Week

6

Monthly Aggregates Retention Period (TOTAL)

Month

· To chec k that the above reorganization is functioning correctly, choose the following path in the navigation bar: Performance Databas e →

Workload Collector Datab ase →

Reorganization →


Log . You can view the corresp onding error logs, broken down

by time periods. The collector runs system-wide once a day. Workload Monitor start page

Displaying and Deleting Contents of the Workload Collector Database Use The data of the workload collector database, which contains the workload data for the system is automatically reorganized and deleted after the set retention periods are exceeded (see Controlling and Monitoring the Generation of Statistics Data ). However, if problems occur during the reorganization of the statistics data, you can also display the corresponding content and delete it manually.

Activities To disp lay or delete the contents of the workload collec tor databas e, start the workload monitor (transac tion ST03 ), and ensure that you are in the Expert user mode.

Displaying Contents of the Workload Collector Database To check the periods, instances, and aggregates for which there is data in the workload collector database, choose the following path in the navigation bar: Collector and Performance DB →

Performance Database →

Workload Collector Database →

Contents .

The period type specifies the type of aggregate. Period Type

Aggregate

D

Daily aggregate

W

Weekly aggregate

M

Monthly aggregate

Deleting Contents of the Workload Collector Database To delete data from the workload collector databas e for indiv idual periods, instanc es, and aggregates, choose the following path in the navigation bar:


Collector

Page 24 of 27

and Performance DB →

Performance Databas e →

Workload Collector Databas e →

Contents . Select the desired data and choose

( Delete ).

Deleting Statistics Records in the File System of the Instance You can also directly delete the unaggregated statistics files in the file system of the individual instances. To do this, choose the following path in the navigation bar:

Collector and Performance Database →

Statistics Records & File →

Delete File , and select the desired instance by double-click ing it. The

system asks whether you want to delete the dialog step or the application statistics file.

If the workload collector is running correctly, both files are automatically deleted when they exceed a certain age (by default, two days). You should therefore only delete the corresponding files manually if there are problems when reading and processing the statistics files. The deletion is performed without aggregation of the data in the workload collector database. After the deletion, this statistics data is also no longer available for detailed analysis. Workload Monitor start page

Displaying and Changing the Scheduling of Performance Collectors Use The workload monitor reads the workload statistics from the workload collector database. Various collectors must run at regular time intervals so that this database always contains current data. This is done using the background job SAP_COLLECTOR_FOR_PERFMONITOR. This job runs once an hour on an instance of the system. The job executes report RSCOLL00, which starts the various collectors in accordance with the entries in table TCOLL.

Note that the collectors in table TCOLL not only form the basis of the data for workload monitor, but are also of importance for other transaction (ST02, ST04, ST06, ST10). This means that the collectors supply not only the workload collector database, but also the monitoring database MONI with values. By changing the entries in TCOLL, you can determine which collectors are to run and how often. Usually, you do not need to change these settings. There are two conceivable situations in which you would need to change the entries in TCOLL. · The workload collec tor databas e or the monitoring database M ONI is too large due to the quantity of collected data. · The collectors themsel ves are caus ing too much workload.

For information about the various collectors, for which the scheduling is determined using the entries in table TCOLL, see SAP Notes 12103 and 127642. Both SAP Notes are regularly updated.

Procedure 1. Start the Workload Monitor by c alling transaction ST03. 2. Ensure that you are in the Exp ert user modesapurl_link_ 00 03 _00 01_0 002. 3. In the navigation bar, choose the entry


Performance Monitor Collector →

Execution Times. The TCOLL List

screen appears. This shows an execution timetable for the various collectors. Each row specifies for a collector on which week days and at which hours of the day the relevant report is executed. For an execution, both conditions (week day and hour) must be fulfilled.

Note that the columns for the week day all hav e the same column heading Day or Weekday . The first column of this typ e stands for Monday, and the columns then run in the normal sequence. The Type column the instances on which the report is to run. Type

Meaning

S

The report is executed exactly once in the system (in the local background work process).

A

The report is executed on every instance.

H

The report is executed once per host.

%

The report is executed on the instance specified in the Application Server field.

4. If you want to change execution times , choose the Table Maintenance button. The Performance Collector: TCOLL maintenance sc reen appears . 5. Each row contains the execution times of a report. - Use seven c haracters in the column Days:M-W-F-S to specify the week days on which the report is to be executed. Enter a character for every weekday, starting with Monday: X (the report is to be executed on the corresponding weekday) or space (the report is not to be executed on the corresponding weekday). - Proceed in the same way in the Hours: 0-2-4-6-8-0 -2-4-6-8-0 -2- column, in which 24 indiv idual charac ters stand for the hours of a day from midnight to 11 p.m. - Sp ecify how often the report is to be exec uted per pass in the column R. - Specify where the report is to be executed in the column S (see the Type column). - You can explicitly sp ecify an instance in which the report is to be executed in the Alternative System column. Workload Monitor start page


Page 25 of 27

Reorganizing and Monitoring the Monitoring Database MONI Use Up to SAP NetWeaver 04 (SAP Web AS 6.40), workload data and various other performance data was stored in the monitoring database MONI in the long term. With SAP NetWeaver 7.0 (SAP Web AS 7.00), the workload statistics that are displayed in transaction ST03 are stored in the workload collector database. However, the monitoring database MONI is still used. It contains, among other things, performance data from the following sources: ● Table call statistics ● Function code statistic s ● Op erating sys tem statistics ● Buffer statistic s ● SQLR summary Although this data is not displayed in the workload monitor, you can still set the reorganization of this data in the monitoring database MO NI in transaction ST03. You can determine the retention period of the existing data in MONI. In this way, you limit the size of the table to optimize the performance of the CCMS monitors. The smaller MONI is, the better the performance of the various CCMS monitors is.

Activities To perform the following functions, start the workload monitor (transac tion ST03) and ensure that you are in the Expert user mode.

Setting Retention Periods for the Performance Database To set the parameters that affect the reorganization of the table MONI, choose the following path in the navigation bar: Performance Database →

Monitoring Database →

Collector and Performance DB

→

Reorganization. The Retention Time for Performance Data screen appears.

This disp lays s ettings for Default Statistics D ata and Other Statistics D ata . The monitoring databas e is therefore divided into two parts, and the data for each of these is retained for different lengths of time. You can see which data belongs to which group in the respective group boxes. During a reorganization run, the system deletes all data whose retention time has expired.

When statistics are deleted through the new retention times used in a reorganization of the monitoring database MONI, these lost statistics cannot be recovered except through the standard techniques for database recovery. We recommend the following retention periods: ● Standard Data This data is very detailed and requires a great deal of storage space. The retention period should therefore be relatively short. Set a value of at least "2" for all three types of statistics, so that the statistics are retained for two days, two weeks, and two months respectively. However, you should use longer retention periods for small and medium-sized SAP sy stems and for testing new SAP releases. ● Other Data This data is less detailed than the standard data and requires significantly less storage space. The retention period can therefore be longer than the retention period for the standard data. Use a retention period of 20 days, 20 weeks, and 14 months. Displaying and Deleting Contents of the Monitoring Database To obtain an overview of the contents of MONI, choose the following path in the navigation bar: Database →

Monitoring Database →

Collector and Performance DB

→

Performance

Contents. The Performance: Contents of Performance Database MONI screen appears.

This shows the rows of the database, where one row is characterized by the following data: ● Data source (such as typ e of data and instanc e) ● Period (day, week, or month) ● Date

The table MO NI is s pecially encrypted and cannot be acces sed using c onventional methods. To delete data, choose the desired command from the Extras → Delete Data menu. Workload Monitor start page

1.2.5 Saving User-Specific Views Use You can save user-specific views in the workload monitor. If, for example, you are mainly interested in the response time analysis of a time profile in the Service Engineer user mode, you can s ave this as the standard view. The next time you call transaction ST03 , the sys tem automatically displays this view. You can save a specific analysis view, with the selected tab p age, task type, and the

ALV layout in a user-specific view.

If you want to save a user-specific v iew, the instance in which you logged on to that view is inc luded in the Expert and Administrator user modes. If you have chosen the Service Engineer mode, the total of all instances ( TOTAL ) is disp layed by default (see also Controlling the Generation


Page 26 of 27

and Retention Period of Statistics Data).

Procedure 1. Start the Workload Monitor by c alling transaction ST03. 2. Choose the instance and/or the period that you want to analyze b y double-cl ick ing. 3. Select the desired analysis v iew, and specify the task types and parameters that you want to display as y our standard view. 4. Choose Save View.

Result The next time you call transaction ST03, the system automatically disp lays the view that you have just s aved. Workload Monitor start page


Page 27 of 27

ST03N Workload Monitor

Recommend Documents