Sizing SAP® Systems Susanne Janssen, Ulrich Marquard
Contents 1
Introduction
.................................................
3
Basic Considerations and Assumptions
Sizing Definition ....................................
3
for Throughput Sizing ............................ 23
The Sizing Principle ................................
3
Advantages and Disadvantages
Business Management and Technology
4
of Throughput Sizing ............................. 23
Goals of This Book .................................
4
Sizing by Reference Installations ........... 23
Target Group and Structure of the
Sizing by Load Tests ...............................
24
Book .......................................................
5
Conclusion .............................................
24
Related Topics ........................................
5
3.4 User and Throughput Sizing Models .................
24
Calculating CPU Requirements .............
24
7
Calculating Memory Requirements .......
25
2.1 Phases of Sizing Projects ...................................
7
Calculating Disk Requirements ..............
26
2.2 Methods for Initial Sizings ................................
8
Frontend Network Requirements .........
27
2
Sizing Methods
...........................................
Hardware Budget Sizings .......................
8
Conclusion for These Approaches .........
27
Advanced Sizing .....................................
10
3.5 Conclusion ........................................................
27
Expert Sizing ..........................................
10
Standard Tools — Even for Experts .......
11
4
29
Analyzing Customer Data ......................
11
4.1 Rule of Thumb/Reality Check ........................... 29
2.3 Sizings Based on Productive Customer Data ....
12
Bottom-Up Method .............................. 30
13
4.2 T-shirt Sizing ...................................................... 30
Basic Analysis for All Production Sizings ....................................................
Sizing Tools
...................................................
Top-Down Method ................................ 30
Resizing ..................................................
13
Categories ..............................................
31
Delta Sizing ............................................
14
Pros and Cons ........................................
31
Upgrade Sizing .......................................
15
4.3 Sizing Formula ...................................................
32
Single-Instance Projects .........................
15
4.4 Offline Questionnaire .......................................
33
2.4 Summary ...........................................................
15
4.5 Summary ...........................................................
33
3
.....................................
17
5
...................................................
35
3.1 Factors That Influence Sizing ............................
Sizing Approaches
17
5.1 Quick Sizer Projects ..........................................
36
3.2 Key Performance Indicators ..............................
18
Creating a Project ..................................
36
Filling Out Sizing Questionnaires ..........
37
3.3 Overview of Different Sizing
Quick Sizer
Approaches .......................................................
20
Determining the Sizing Result ............... 38
Sizing by Users .......................................
20
5.2 Functions ........................................................... 40
Advantages and Disadvantages of
Initial Page .............................................
41
21
Navigation Tree ......................................
41
Sizing by Throughput ............................. 22
Header Bar .............................................
41
User-Based Sizing ..................................
www.sap-press.com
1
Contents
Questionnaires .......................................
43
Step 5: Acquire Information and Apply
Results Page ...........................................
45
the Methods ..........................................
76
5.3 Average and Peak Sizing .................................... 48
Step 6: Analyze First Results and Adapt
5.4 Summary ........................................................... 50
the Methods .......................................... 77 Step 7: Consolidate the Results and
6
Performance Monitors and Traces
......
51
6.1 Operating System Monitor ...............................
52
6.2 Database Monitor .............................................
53
Get Confirmation from Stakeholders .... 77 8.4 Summary ...........................................................
78
6.3 Application Monitor .......................................... 54
9
79
6.4 Single Record Statistics .................................... 56
9.1 Basic Foundations of the SAP Sizing
6.5 Performance Trace .............................................
Sizing Details
...............................................
57
Model ................................................................ 79
6.6 Summary ........................................................... 58
SAP Software Architecture .................... 79 Application Services and Database
7
Sizing Verification
......................................
59
Services .................................................. 80
7.1 Load Tests ..........................................................
59
Modeling CPU Consumption ................
Phase 0: Preparation ..............................
59
Allocating Sufficient Memory
81
Phase 1: Performing Individual
(or: Modeling Physical Memory) ........... 84
Measurements ....................................... 60
Allocating Sufficient Disk I/O
Phase 2: Analyzing the Transaction
Capabilities (or: Modeling Disk I/O
Design in Single Mode .......................... 60
Requirements) ....................................... 86
Phase 3: Load Tests in Multi-User
Modeling Network Bandwidth .............. 86
Mode .....................................................
61
Measuring Resource Consumption ....... 88
7.2 Verification via Support Services ....................... 63
Benchmark Results ................................ 88
SAP GoingLive Check ............................ 63
Results from a Java Benchmark ............. 90
SAP EarlyWatch Check .......................... 67
9.2 SAP Application Performance Standard ............
SAP GoingLive Functional Upgrade
9.3 Performing Sizing Measurements ..................... 94 Step 1: Define the Test Case ................. 94
Check ..................................................... 68 7.3 Summary ...........................................................
92
Step 2: Identify the Test System ............ 95
69
Step 3: Create the Test Case in the
8
Executing Sizing Projects
........................
71
Test System ............................................ 95
8.1 Before the Sizing Project Begins .......................
71
Step 4: Measure the Sizing KPIs ............ 96
Chicken or the Egg? ...............................
71
Step 5: Create Sizing Guidelines Based
Project Scope .........................................
71
on the Measurements ........................... 98
Stakeholders in a Sizing Project ............. 72
9.4 Summary ........................................................... 99
Definition of a Sizing Project ................. 72 8.2 Project Team ...................................................... 73 8.3 Methodical Procedure ......................................
A
Step 1: Define Project Contents and Goals ......................................................
74
Step 2: Determine Performance-Critical Processes ................................................
10 Summary and Outlook
............................ 101
74
Frequently Asked Questions
................. 103
A.1 Sizing Approaches ............................................. 103 A.2 Quick Sizer ........................................................ 104
75
A.3 Sizing Projects ................................................... 104
75
B
Step 3: Decide on the Approaches and Methods to Be Used ..............................
Literature and Links
.................................. 105
Step 4: Define Milestones and Prepare a Detailed Schedule ...............................
2
© Galileo Press 2007. All rights reserved.
76
Index
............................................................... 107
2 Sizing Methods
Sizing projects are carried out at very different stages of
sizing). Moreover, we are using several custom develop-
an SAP project. They represent an iterative process that
ments. How should we carry out a sizing project?”
depends closely on the amount of information that is
This question refers to a specific component in
available to you at a certain point in time to make reliable
accounting and is therefore more detailed. Perhaps
statements on the actual hardware requirements.
this customer has already carried out sizing projects
Accordingly, in each sizing project, you will often face
for other SAP applications and wants to perform
new situations that you must react to with different meth-
another one for this particular application. In addi-
ods and, consequently, using different tools. This chapter
tion, the customer wants to know how sizing can be
focuses on these different methods.
done for proprietary developments. 왘
2.1
“We are planning to consolidate our seven data centers into one. Can we simply add up existing sizings?”
Phases of Sizing Projects
This question (which comes from an existing SAP
SAP regularly receives information requests like the fol-
customer) refers to a system consolidation process
lowing:
in which additional hardware requirements must be
왘
“We are a large customer in the consumer goods indus-
taken into account if the different existing systems
try with 30,000 business partners and 60,000 sales
are combined. System consolidation and single-
orders containing 50 line items per month. How much
instance concepts, which are used to check whether
hardware do we need for our SAP application?”
all systems can be globally integrated with one data-
This is a rather general question. The customer needs information about hardware for a first estimate. The
base, are currently red-hot issues with our customers. 왘
“We are currently running Release SAP R/3 4.6C and
question itself does not indicate why this is a large
want to upgrade to SAP ERP 6.0. What are the upgrade
customer. Perhaps the customer is only looking for a
factors?”
partial solution since the volumes mentioned indicate
This customer uses a specific release that he wants
that this customer is a large medium-sized company.
to upgrade across multiple releases in one step and
The business partners represent master data and are
therefore wants to know if new hardware needs to be
not yet relevant to sizing because they do not gener-
purchased for that.
ate any load during live operation. In contrast, the
왘
sales orders and sales order items are much more
By further analyzing these kinds of requests, we ulti-
critical to CPU sizing since they represent transac-
mately get to the different phases in which you can per-
tion data. In terms of revenue, an average of 2,000
form sizing projects (see Table 2.1). The informational
sales orders per day is quite considerable; however,
value of the sizing project can vary, depending on the
from the point of view of software, this is not a high
different phases. In addition, you should note that not
throughput volume. SAP has several customers who
all the phases described in Table 2.1 have to occur in an
process more than a million sales order items per day.
SAP project.
“We can’t find any guidelines for the FIN-FSCM-TRN component in your sizing area (http://service.sap.com/
Thus, if the system GoLive is still way down the road and you — as a customer — are not yet very familiar with
www.sap-press.com
7
2 Sizing Methods
Phase
Point in Time
Description
Orientation phase (Phase A)
18 to 12 months prior to GoLive
You familiarize yourself with the software functionality and want to know what the range of expenses is for the new hardware. Accordingly, you will certainly know which processes you want to map using the software, and you also know the approximate amount of data that is supposed to be processed. However, you are not familiar with the SAP jargon, nor are you interested in specific releases.
Blueprint phase (Phase B)
12 to 6 months prior to GoLive
The first business blueprints have been created, and now you need reliable information on the scope of hardware you have to order because you must make sure you meet all your deadlines. You know how to implement the relevant processes, have become more familiar with SAP products and SAP terminology, and know which release you want to use.
Implementation phase (Phase C)
6 to 0 months prior to GoLive
You have ordered the hardware or are just about to do so, and you want to be absolutely sure that sizing is correct. For example, you are able to measure core processes using the performance monitors.
Consolidation phase (Phase D)
System is operational
The system is operational and is supposed to be consolidated. Region 1, for instance, has gone live with a specific software, and Region 2 is now supposed to go live on the same system.
Extension phase (Phase E)
System is operational
The system is operational and you want to add new functions. For example, your live system runs the SAP ERP applications, and you want to add CRM applications now.
Upgrade phase (Phase F)
System is operational
The system is operational and you want to perform an upgrade. For example, the system runs on SAP R/3 Enterprise and you want to upgrade it to SAP ERP 6.0.
Table 2.1 Phases in Which Sizing Can Be Performed
the software, you will probably have only basic informa-
ings in phases B and C, resizings in phase D, delta sizings
tion on how you are going to use it so that you can at
in phase E, and upgrade sizings in phase F.
least make a rough estimate of the costs involved. During the course of the implementation project, you can refine your initial assumptions by using standard sizing rules in order to take a closer look at the critical issues.
2.2 Methods for Initial Sizings Initial sizings are sizings that refer to new installations,
If an installation’s complexity differs too much from
that is, installations in which you use SAP software for
the standard, you can, for instance, measure customer
the first time. That is also the case if, for instance, you
processes in order to create customer-specific sizings.
want to use SAP SRM for the first time while SAP ERP
All these different phases require different sizing meth-
is already running in your company’s production system
ods. In this context, we generally distinguish between ini-
— at least the sizing for SAP SRM will be considered as
tial and production sizings. Figure 2.1 provides an over-
being initial.
view of the available sizing methods, with initial sizings
Depending on the project phases, we differentiate ini-
being shown in the upper section and production siz-
tial sizings into hardware budget sizings (budget sizings
ings in the lower one. Expert sizing is marked as a hybrid
for short), advanced sizings, and expert sizings. Usually,
because under certain circumstances, some processes can
budget sizings and advanced sizings are based on tools,
be mapped using standard methods while, at the same
whereas expert sizings are a mixture of tools and addi-
time, customer-specific data can be analyzed.
tional rules or measurements.
The following sections describe the characteristics of these different sizings in greater detail. At this point it is
Hardware Budget Sizings
important to know that sizings can be performed within
The main characteristic of budget sizings is that they
several phases of a project: Sizing is an iterative process.
don’t require much information from the customer and
Budget sizing, for example, could be done in phases A
they contain many assumptions (i.e., values provided by
and B, advanced sizings in phases A through C, expert siz-
SAP based on experience). For example, if the only infor-
8
© Galileo Press 2007. All rights reserved.
2.2 Methods for Initial Sizings
Hardware Budget Sizing Smaller Companies
Advanced Sizing
Expert Sizing
GoLive
Large/Complex Projects
Medium to Large Companies
Very simple algorithms
Throughput estimates
Additional guidelines
Assumptions, likelihoods
Questionnaires, formulas
Custom calculations
Level setting of project
Usage of standard tools
Analysis of custom coding
Risk identification
Focus on core business processes
Custom sizing guidelines
Initial Sizings
Resizing
Delta Sizing
Upgrade Sizing
All Projects
All Projects
All Projects
SAP system monitors
SAP system monitors
SAP system monitors
Goal: Extend an existing system by load (e.g., by volume, 100 additional users who will do the same as the current productive ones)
Goal: Extend an existing system by functions (by different functions, e.g., you are live with CRM and want to add SCM)
SAP Notes
Goal: Upgrade SAP software
Post GoLive Sizings Figure 2.1 Overview of Sizing Approaches and Methods 1
mation you have is that 100 users will use SAP CRM, but
Budget Sizings Help in Estimating the Entire Size
you don’t know the other applications they will use and
Let’s suppose a budget sizing determines 4,000 SAPS
what will be their average activity, you can certainly per-
(SAP Application Performance Standard1). Currently,
form the sizing, but in the long run, the informational
4,000 SAPS correspond more or less to a dual-core
value provided by the result of the sizing process will be
machine (server) with two processors, which has a list
too restricted.
price of $15,000. Now you can make up your mind
For this reason, budget sizings are usually performed
whether it makes sense to tackle a rather intensive siz-
way ahead of the GoLive phase (most of the time in
ing process or whether you want to take one of the
Phase A) if the goal is to estimate the approximate scope
following two risks:
of hardware.
왘
For budget sizings, you can use the user-based sizing
Result Is Too High This means the server will not be fully utilized dur-
function in SAP’s Quick Sizer (see Chapter 5, Quick Sizer).
ing live operations. A result that is too high often
Alternatively, you can use T-shirt sizings (see Section 4.2,
occurs because the initial estimates are usually too
T-shirt Sizing), which have the advantage of requiring you
conservative.
to answer only a few questions. Of course, the disadvan-
왘
Result Is Too Low
tage is that the rough categorization into S through XL
This means that you must buy additional hard-
provides only limited informational value. Occasionally,
ware. In this case, the question is whether you can
such sizings can be sufficient, depending on the specific
afford to use the wrong assumptions. Let’s sup-
situation.
pose your initial estimate is wrong by 100%. You
For this reason, it makes a lot of sense to compare the time and effort you want to invest into a sizing project with the potential hardware costs.
1 See Section 9.2, SAP Application Performance Standard, for a more detailed description of SAPS.
www.sap-press.com
9
2 Sizing Methods
as the size of these objects. If you have times of peak
would then have to pay (in the above example)
load, you can, of course, specify them.
an additional $15,000 – $20,000 for a correspondingly bigger server. There are some customers for whom expenses in this range are critical, since the implementation of a new production server also involves the purchase of new quality assurance systems and testing landscapes.
왘
Throughput-based sizing enables you to determine in greater detail in which areas and at what time the CPU peak load occurs (for example, in the week before Christmas or New Year’s). Especially with regard to background-oriented processes such as those relevant to controlling or year-end settlements, this information is critical and cannot be taken care
Advanced Sizing
of by user-based sizing.
If you’re in a situation in which there’s a high risk of misjudging the requirements by several 100 percents, you
The drawback of advanced sizing is that you have to
should refine your budget sizing by using what is referred
familiarize yourself with the core business processes in
to as advanced sizing. For example, if the range of CPU
order to obtain the appropriate information from the
power you’re dealing with is between 8 and 16 cores, a
user departments for the Quick Sizer questionnaire. This
more detailed sizing makes a lot of sense because it pro-
certainly takes more time than asking for the number of
vides a higher degree of reliability. To do that, you can use
users (as is done, for instance, in a budget sizing process),
additional functions of Quick Sizer, such as its through-
but it is much more accurate.
put-based functionality, which allows you to determine
Note that advanced sizing is still a tool-based process.
the CPU load on average as well as by peak load (see Sec-
An “XL” category in Quick Sizer represents a large cat-
tion 5.3, Average and Peak Sizing).
egory in the tool-controlled area, but not necessarily in
Usually, advanced sizing occurs in phases B and C. In
the entire sizing context. For example, in Quick Sizer, the
these phases, the first business blueprints have already
largest category (“XXL”) starts at 30,000 SAPS. A number
been created so that important and sizing-relevant infor-
of large customers operate on 40,000 to 100,000 SAPS;
mation about the business software applications is avail-
a few other customers operate in the range of 300,000
able to you. This information could include, for instance,
SAPS and higher.
a PC vendor’s decision about which important materials are imperative that an availability check be performed
Expert Sizing
for (processors, for example). An availability check locks
For ranges of 30,000 SAPS and higher, SAP therefore rec-
an object and can become a performance bottleneck
ommends that its customers not rely exclusively on one
because all other requests have to wait until the object is
sizing tool but rather that they analyze the core processes
released again.
and, above all, the customer processes in great detail via
Thus, in an advanced sizing process, you focus more
expert sizing.
on the core business processes. Quick Sizer is able to map
This method is particularly suited for complex busi-
the key processes of the SAP Business Suite and tries to
ness transactions, in-house developments, and large-
break down the complex business scenarios into the most
scale installations. Complex business transactions may
important transactions and objects. In addition, Quick
also occur in smaller installations, such as in the supply
Sizer provides the option to fine-tune the CPU sizing in
chain or in retailing systems. Global installations, which
that it distinguishes between the average CPU utilization
are not only defined by their size, are also eligible can-
(average sizing) and the utilization at peak times (peak siz-
didates for expert sizing because of the time differences
ing):
that must be taken into account.
왘
For processor requirements, you can perform an average sizing in such a way that you specify the number of objects that are processed per year as well
10
© Galileo Press 2007. All rights reserved.
To be able to perform an expert sizing process, you must have:
2.2 Methods for Initial Sizings
왘
Identified all processes that are critical for performance.
왘
Used standard tools for the core processes.
왘
Determined the performance-critical areas in which your processes deviate from the standard.
Simplified Example of Expert Sizing A company uses SAP CRM applications to enter sales orders and uses SAP ERP for sales order fulfillment and HR. The sales order processing functions in the ERP system have been custom-coded. For this reason, a mixed approach is used in expert
Expert sizings are performed just before the system GoLive, that is, when the functionality has been clearly defined and perhaps even been implemented. In most cases, expert sizing represents an iteration on a previously performed budget or advanced sizing so that you can use the data of these previous processes as a basis and simplify it, if necessary.
sizing in such a way that core processes are mapped through the standard as much as possible, while the other processes are approached step by step: 왘
standard sizing for sales order entry in SAP CRM. 왘
ERP system, a certain amount of extra capacity is
mixture of standard sizing and performance tools, and on
added to the sending system, that is, SAP CRM,
the other, of additional procedures and analyses. We can 왘
The full utilization of the sizing tools’ functionality (in
according to the corresponding sizing rules. 왘
that the orders are transferred through an inter-
requirements at least in part.
face does not negatively affect the performance
The analysis and performance monitoring of core
of the ERP system (on the contrary, it has, rather,
processes in the customer system.
a positive effect because there is no user interaction). This sizing represents the basic structure of
The following sections provide an overview of how you can use standard tools in expert sizing to obtain useful information, at least about parts of your system.
the ERP sizing. 왘
cessing will be, it performs performance measurements that show that, because of the extension
Whenever you have identified business transactions as
made in the customer system, the same process
being close to the standard, you can use Quick Sizer (see
that was mapped in Quick Sizer now needs more
Chapter 5). That is, you can use Quick Sizer for partial Another option for using Quick Sizer in expert sizing is that you can use it for optimizing process flows from the point of view of sizing. For example, if you use overlapping, performance-critical process chains, you can use the 24-hour load profile provided by Quick Sizer to ascertain
Because the company does not know up front what the impact of extending the sales order pro-
Standard Tools — Even for Experts
sizings.
The sizing of SAP ERP is mapped in Quick Sizer on the basis of the total number of orders. The fact
particular, Quick Sizer’s) so that they meet specific 왘
Because the sales orders that have been entered in the CRM system are further processed in the
Basically, this method consists, on the one hand, of a
roughly subdivide these two parts into:
First the company uses Quick Sizer to create a
resources. 왘
The customer is now able to increase the ERP result for sales order processing by 30% in such a way that the customer multiplies the Quick Sizer result by a factor of 1.3. Other results (for instance, in HR) are not affected by this.
whether it is possible to perform moves (see also Section 5.3, Average and Peak Sizing). Quick Sizer enables you to
Analyzing Customer Data
map and document additional loads which, for example,
One of the most important tasks of expert sizing consists
have been caused by custom coding.
of analyzing specific customer processes. Typical cases in which it makes sense to analyze the performance figures on the basis of custom data because of the strong inherent customer-specific nature include the following:
www.sap-press.com
11
2 Sizing Methods
Variant configuration that evaluates complex object
왘
Sizing Measurement Versus Performance Analysis
dependencies: Its runtime can hardly be anticipated
Note that sizing measurements reflect only the actual
in the standard, if at all.
status. Based on sizing measurements, you can deter-
Each custom extension.
왘
mine whether a business transaction is scalable. In this context, scalability means that the resource con-
To analyze customer data, the following two methods are
sumption increases linearly with the number or size
available: single-user analysis and the load test.
of the processed projects. If a process is not scalable,
Single-user Analysis
왘
you must analyze and resolve the problem in a perfor-
Single-user analysis is based on a relatively simple
mance subproject.
principle: As soon as integration tests can be performed (i.e., when business processes can be functionally mapped in a system), you use the standard
The advantages of expert sizing over other sizing meth-
performance monitors of the SAP system to mea-
ods are found in the higher degree of accuracy and reli-
sure the CPU time, memory consumption, or data-
ability of the information. If you manage a sizing project
base growth on your hard disk, depending on your
for a complex or large customer, you should definitely
requirements. You can then use this data in a rule of
consider aspects from expert sizing, even though the col-
three to create the sizing forecast.
lection and analysis of the information takes more time.
Table 2.2 provides an overview of the procedure to be applied in a single-user analysis, from defining an appropriate test case to applying the customer-specific sizing rule. Chapter 6, Performance Monitors and
2.3 Sizings Based on Productive Customer Data
Traces, contains detailed information on sizing-based
Sizing is an iterative process — that is, even operational
performance measurements.
installations can be subject to change processes that affect the resource requirements, as the following exam-
Step
Description
1
Define test case
2
Identify test system
3
Create test case in test system
4
Measure sizing KPIs
5
Implement measurement results in sizing method
system (for example, by installing a CRM system on a
6
Apply sizing rule
server that already hosts an ERP system).
Table 2.2 Steps in Creating a Sizing Rule 왘
ples will show: 왘
scape (for example, by merging all your international subsidiaries on one server). 왘
왘
You want to add additional functions to an existing
You want to upgrade Release X to Release Y.
Load Test
All these situations can affect the hardware and require
Load tests are occasionally used in the context of
a more or less comprehensive sizing project. The major
expert sizings and make sense when a single-user
advantage of sizings that are based on a production sys-
analysis does not provide sufficient information about
tem is that you can use your own data and settings as a
the locking procedure or memory requirements.
basis. In other words, you do not need to rely on assump-
In the sizing environment, load tests have a hybrid
tions made by SAP.
nature: On the one hand, you can use them as a siz-
Regarding production sizings, we distinguish between
ing tool. On the other hand, you can use them to
the following three methods, which pursue different
verify sizing results. Because customers usually use
goals:
them to verify sizing results, you can find detailed
왘
information on them in Section 7.1, Load Tests.
12
You want to consolidate your existing system land-
© Galileo Press 2007. All rights reserved.
Resizing In a resizing project, the throughput or user volume
2.3 Sizings Based on Productive Customer Data
왘
왘
changes, but not the processes (or customizing or
it can also be projects or calls. Alternatively, you can also
parameter settings, and so on).
use the number of activities or sales orders, depending,
Delta Sizing
on the one hand, on which unit is best suited to reflect
In a delta sizing project, you add new functionality.
the respective business activity, but also, on the other, on
Upgrade Sizing
how easily it can be determined.
An upgrade sizing involves a change of the SAP release.
Sample Analysis of a Production System The following example forms the basis for the descrip-
Common to all these sizing methods is that you must first
tion of individual sizing methods. A customer uses
analyze the status of the existing system before you can
strategic procurement in the SRM environment. The
plan the new hardware requirements.
analysis of the current utilization provides the following result:
Production System Sizings Versus Quick Sizer
왘
CPU
The unbeatable advantage of sizing on the basis of pro-
Utilization of the database server is 34%; that of
duction data is that you can take your own data, pro-
the two application servers is 56%.
cesses, and settings into account. Quick Sizer has been
왘
designed for new installations and contains assump-
213GB out of 512GB are occupied with a monthly
tions about the productive operation. For this reason, we recommend Quick Sizer for initial sizings only.
Database growth of 7GB.
왘
Memory 26GB out of 32GB are being consumed.
Basic Analysis for All Production Sizings
By using a system monitor, the customer has found
For all production sizings, you must first identify the uti-
out that approximately 1,254 named users out of a
lization of the sizing-relevant components in the exist-
total of 1,567 have been active during the period ana-
ing system. Using the appropriate monitors, which are
lyzed. Based on this information, you can now deter-
described in detail in Chapter 6, you can determine the
mine whether the existing hardware is sufficient or
following information:
whether it must be extended.
왘
왘
CPU Utilization What is the actual utilization of the CPU? Can the
Resizing
existing hardware compensate for the future load?
A basic prerequisite for resizing is that only the through-
Here, you must distinguish between the utilization of
put and user volumes can change, but not the function-
the application server and that in the database.
ality. Based on the current load situation and the new
Memory Consumption
information, you can easily determine future require-
How much room for maneuver do you have regard-
ments using a rule of three.
ing the memory requirement: Will it increase or stay 왘
Typical resizings occur in system consolidations or in
the same?
what is referred to as phased rollouts, in which customers
Database Space
install new software in different phases in their business
Take a look at the 30 biggest tables and indexes, and
units or international subsidiaries.
make a note: How quickly did they grow during the last several months?
Resizing a Production System Based on the above example (see previous box, Sam-
Once you have determined the current utilization or the
ple Analysis of a Production System), a resizing could
database growth and the increasing memory require-
look as follows: You want to add another 200 named
ments using the various vendor-specific monitors or the
users to the 1,567 existing ones. We assume that the
SAP monitors, you should relate this information to a
ratio between named users and active users is identical
simple business key figure. Usually this is the users, but
www.sap-press.com
13
2 Sizing Methods
among the new users and that they will do the same as the existing users, so that we can make the following calculations: 왘
Active Users The ratio between 200 and 1,567 is 12%, which means that the number of active users will probably increase by 12%.
왘
CPU Database Server 34% + 12% corresponds to 34% × 1.12 = 38.1% A utilization of 38% is sufficient for a database server. Many customers plan a target utilization of 25% to 50% for the database server.
왘
CPU Application Server 56% + 12% corresponds to 56% × 1.12 = 62.7% The application servers can absorb a utilization of 62.7% quite well. However, many customers plan a target utilization of 30% to 50% for the application servers, which is why an extension is at least
value, which can be specified by the hardware vendors at any time and for each release. Based on this information (available SAPS, software release, CPU utilization, new SAPS), you can easily calculate whether the hardware will be sufficient by using a rule of three. Delta Sizing of a Production System The above customer (see previous box, Resizing a Production System) has created a sizing for a new application. According to the sizing, the application will require 1,200 SAPS (240 database SAPS and 960 application SAPS). What needs to be done now is easy: The SAPS values must be added up, and the target utilization must be determined. The existing hardware is evaluated as follows: 왘
tions: 2,800 SAPS each 왘
3,700 SAPS at the application level (i.e., 66% of
Main Memory
5,600 SAPS)
26GB (out of 32): 26GB × 1.12 ~= 29GB 29GB out of 32GB of allocated memory might be a bit tight. It is therefore advisable to extend the memory. 왘
Database Space 7GB of growth corresponds to 7GB × 1.12 = 8GB per month
For the database, this would mean the following: 1,360 SAPS + 240 SAPS = 1,600 SAPS — which represents a future utilization of 40%. The application servers reach 4,660 SAPS and a utilization of 83%, which could lead the customer to the conclusion that it would make sense to add another application server. If you have followed the above descriptions of tools
Currently, 312GB out of 512GB are being used. If the database grows by 96GB (8GB × 12 months) per year, bottlenecks can occur in a very short time. Thus, the disk space should be extended.
The current net SAPS consumption for the database is 1,360 SAPS (i.e., 34% of 4,000 SAPS) and
conceivable here. 왘
Database server: 4,000 SAPS; the two applica-
and methods closely, you will have noticed that SAP calculates the standard sizings with a target utilization of 65% and you should therefore only use net amounts. However, you should also take into account that the delta is based on standard assumptions as
Delta Sizing
well, and the 65% factor could be a useful buffer. But if you want to base your calculations on net
Because delta sizing can be performed only when new functions are added to an existing software application,
amounts, you can do so as follows:
the procedure is similar to that of initial sizing, the only
왘
The net requirement of the new application is 780
difference being that the sizing results are applied to the
SAPS (1,200 SAPS × 0.65 target utilization). 160
current utilization.
SAPS out of the 780 SAPS are allocated to the database, 620 SAPS to the application level.
However, there is one special characteristic you should be aware of: The SAP’s standard sizing rules specify the
왘
Consequently, this means that we can expect a
CPU requirements in terms of SAPS and a target utili-
growth of approximately 10% for the database
zation of 65%. As we will demonstrate in Section 9.2,
and approximately 20% on the application side.
SAP Application Performance Standard, each hardware configuration in the SAP environment has its own SAPS
14
© Galileo Press 2007. All rights reserved.
2.4 Summary
Upgrade Sizing
Single-Instance Projects
In upgrade projects, customers usually implement numer-
From the point of view of sizing, the majority of single-
ous changes, which include the SAP software, database,
instance projects in which companies change from a best-
operating system, and an exchange of hardware. It often
of-breed strategy2 to a single-instance strategy (one soft-
happens that the configuration and parameter settings are
ware vendor, all data in one system) represent a mixture
changed as well. All this can have an impact on the num-
of resizing and delta sizing, sometimes also upgrade siz-
ber of work processes, buffer settings, or other things.1
ing. Note that an upgrade sizing must be performed first
Upgrade sizing refers to the additional requirements
to make sure that identical conditions apply.
of SAP software. SAP uses regression tests to check the resource consumption of the most important transactions
Considerations in the Context of a
and to create a delta. This information is made available
Single-Instance Study
to all customers in SAP Notes, such as SAP Note 901070,
A customer uses several SAP and legacy systems in
which describes the resource consumption between
Europe. This customer now wants to consolidate its
SAP ERP Core Component (SAP ECC) 5.0 and SAP ERP
European and American systems. Consequently, this
6.0. The SAP Notes provide information about the delta
means the following:
regarding the number of database calls, CPU require-
왘
If the SAP software has different release versions, an upgrade sizing must be performed first. The rel-
ments, memory requirements, and database space.
evant factors will be upgraded so that all systems
Because these SAP Notes provide standardized infor-
have the same version.
mation about different transactions, they carry the risk of 왘
you currently using a transaction that is counterbalanced
The next step involves resizing the SAP software based on the same release version; that is, the cur-
by other transactions.
rent consumptions of existing SAP systems must Sample Upgrade Sizing
be analyzed and totaled.
A (fictitious) SAP Note on delta resource consumption
왘
Finally, a delta sizing must be performed for the
states that the resource consumption in the memory
legacy software. Ultimately, the additional require-
increases by 5% on average. Transactions A and F do
ments for the new software are added to the exist-
not show any additional consumption, whereas Trans-
ing load.
action G consumes an additional 30%. The CPU and database consumptions remain unchanged. If you — as the customer — now use Transaction G
2.4 Summary
extensively, this could cause problems when calculating the main memory. The best thing to do is to calcu-
Because SAP software shows a high degree of scalabil-
late a best case and a worst case.
ity, you can consider a linear change in consumption as
왘
왘
Memory (Best Case)
a given fact. The same applies to hardware: If you want
26GB (out of 32): 26GB × 1.05 ~= 27.3GB
to extend the processing power of application servers,
Memory (Worst Case)
you can add more servers, replace the CPU, or add more
26GB × 30% = 33.8GB
CPUs, depending on your specific production model.
Probably, the future memory requirement will be within that range.
However, a new application server affects the database’s memory requirements because it involves the addition of new database users. A higher volume generally means an increase in read and write activities, which,
1
Since this is a very complex subject, SAP provides the SAP GoingLive Functional Upgrade Check service as part of the standard service coverage (see also Section 7.2, Verification via Support Services). The SAP GoingLive Functional Upgrade Check includes a sizing process.
in turn, may have an impact on the disk subsystem. 2
In a best-of-breed strategy, you always choose the product from the best vendor for each (technological) area. The different products are then connected with each other via interfaces.
www.sap-press.com
15
2 Sizing Methods
The sizing method used essentially depends on the initial
certain limitations. These limitations primarily depend on
position you’re in. Basically, there are different methods
the way in which business processes and the associated
for an initial sizing, which can be mapped via standard
application software interact with each other. For this
tools, and for a production sizing, which uses production
reason, the following chapter, Sizing Approaches (Chap-
data as a basis for forecasting.
ter 3), describes how you can convert user-based and
In this chapter, we have mentioned several times that although sizing tools are very useful, they are subject to
16
Š Galileo Press 2007. All rights reserved.
throughput-based sizings into algorithms, and discusses the pros and cons of different sizing approaches.
Index
2-tier implementation 47
Computing Center Management System
3-tier implementation 47 80/20 rule 35
(CCMS) 51, 65, 68 Concurrent user 21 Consolidation phase 8
A
Core 18
A/P (Average and Peak) 44 Active user 21 Advanced sizing 10 Analysis of customer data 11 of customer processes 11 performance monitor 51 transaction design 60 Application monitor (ST07) 54 Application profile 71
Core process business 10, 65, 75 Core team 73
Extension phase 8
F Frontend network 19, 21, 57
G Garbage in, garbage out 32, 76
CPU 3, 15, 18, 66
GLC 씮 SAP GoingLive Check (GLC)
CPU load 24
GoLive 8
CPU time 3, 12, 23, 30, 57 CPU utilization 13
H
Custom development 8
Hardware budget planning 4, 71
Customer data
Hardware budget sizing 8
analyzing 11
Hardware costs 4, 71
Customer reference installation 23
Hardware vendor 35, 42, 71
Customizing 13, 17, 65
high-volume load tests 24
Average CPU sizing 49
D
I
Average load 48
Database 18, 39, 53
I/O (input/output)
Application server 14, 19, 46, 55 Application team 73 Archiving object 45
Database monitor 53
B
Database server 13
Baseline test 62 Benchmark result 30, 36 BI Accelerator 씮 Business Intelligence Accelerator Blank installation 46, 47 Blueprint phase 8 Business Intelligence Accelerator (BI Accelerator) 18
Database space 13, 39 Data Dictionary 58
Implementation phase 8, 71, 76
Disk I/O operations 19, 39
Implementation project 27, 71, 74
Disk space
Initial sizing 8, 63, 75
database 14, 39, 47 DPU 씮 Logical Deployment Unit (DPU)
E
Chief Process Innovation Officer (CPIO) 4 Coding custom 11, 59, 62
3-tier 47
Disk growth 53
Cache 27
Chief Information Officer (CIO) 4, 74
2-tier 47
Disclaimer 42
C System (CCMS)
(IAS) Implementation
Delta sizing 13, 14
Dual-stack installation 26
CCMS 씮 Computing Center Management
per second 39 IAS 씮 International Accounting Standard
Input error 41, 43 International Accounting Standard (IAS) 33 IT team 73
Employee Self-Services 75
J
Evaluation phase 74
Java-based
EWC 씮 SAP EarlyWatch Check (EWC) Expert sizing 10, 29, 51, 76
application 25, 47 Java Virtual Machine (JVM) 57
Expressiveness of sizing project 7
www.sap-press.com
107
Index
K Key performance indicator 12, 17, 31
Physical memory 18
Single record statistics (STAD) 56
Processor 18
Sizing
Product availability 5
approach 17
L
Product Availability Matrix (PAM) 5, 18
by throughput 22
Production sizing 8, 12, 53, 67
by users 4, 20
Landscaping 6, 72, 78
Programming guidelines 30
definition 3
Latency 19
Project team 73
expressiveness 7
liveCache 18 Load test 12, 59 analysis 60 multi-user mode 61
formula 32
Q
informational value 31 initial 8
Quick Sizer 35
methods 7, 8, 11, 16, 27
average CPU sizing 37
performing 61
measurement 12
CPU peak sizing 45, 48
planning 59
object 45
design principles 35
toolkit 24
principle 3
documentation 36, 41, 42
tools 60
production 8, 51, 63
functions 36, 40
Local area network (LAN) 17, 73
production sizing 13
navigation 41
Logged-on user 21
result 40, 46
peak CPU sizing 37
Logical Deployment Unit (DPU) 46
scope 20
project 36, 41, 47
M
questionnaires 37, 41, 47
Master data sizing 22, 26
Save function 41
Maximum Extended Memory in Transac-
sizing element 38, 44, 47, 48
throughput-based 38, 44 tool 11, 29, 51
result 38
tion 57
user-based 20, 38 verification 59, 63 Sizing project 71
R
application team 73
Reference database 23
definition 72
reference installations 23
documentation 74, 77
Residence time 19, 27
N
execution 71, 74
Resizing 12, 13, 63
goal 74
Resource consumption 15, 24
Named user 20
IT team 73
Rule of thumb 29
planning 71
S
procedure 74
Memory consumption 13 Memory requirement 40, 52, 56, 57 Methods 7 Minimum requirement 5
core team 73
Network load 19 Network traffic 32 Non-disclosure policy 23
O
project scope 71 stakeholders 72
SAP Application Performance Standard
success factors 71
(SAPS) 10, 38, 40, 50 SAP EarlyWatch Check (EWC) 67
Software architecture 31
Offline questionnaire 33
SAP GoingLive Check (GLC) 63
Status 42
Online Analytical Processing 29
SAP GoingLive Functional Upgrade Check
System consolidation 13, 15
Operating system monitor 52 Orientation phase 8
15, 63, 68
System GoLive 63, 67
SAP NetWeaver
System integration 65
P
Business Intelligence 21, 40, 58 Portal 21, 44
T
PAM 씎 Product Availability Matrix (PAM)
Process Integration 4
T-shirt sizing 9, 30
Peak load 48, 49
SAPS 40
Target utilization 14, 50
Peak sizing 49
SAP Solution Manager 64
Throughput-based CPU sizing 45
Performance analysis (ST30) 12
SAP Standard Application Benchmarks
Throughput sizing 22, 23
Performance monitor 51
23
Throughput sizing model 23
Performance trace (ST05) 51, 57
Scalability 3, 61
Throughput volume 7
Phased rollout 13
Sessions 21
Total cost of ownership (TCO) 4
Phases
Single-instance project 15
Trace tool 51, 57
Single-user analysis 12
Transaction DB02 53
of sizing projects 7
108 Š Galileo Press 2007. All rights reserved.
Index
Transaction ST05 57
active 21
Transaction ST06 52
concurrent 21
Transaction ST07 54
interaction step 52, 57
Transaction STAD 56
logged-on 21
U Upgrade phase 8 Upgrade sizing 13, 15, 63, 67, 75
named 20 User-based sizing 20, 38 User interaction step 57
Usage type 47
V
User
Verification 77
of sizings 59, 63
W Wide area network (WAN) 17, 73 Work days in Quick Sizer 42
www.sap-press.com
109