You are on page 1of 33

PERFORMANCE OPTIMIZATION 101

CHANGES TO YOUR IBM COGNOS SERVERS FOR OPTIMAL PERFORMANCE

TODAYS AGENDA
Introduction
System Performance Tuning
Changing dispatcher settings

Improving sorted report performance


Scalability horizontal vs. vertical scaling
Infrastructure issues
Next steps advanced topics

Copyright 2014 Senturus, Inc. All Rights Reserved

SENTURUS INTRODUCTION
WHO WE ARE

SENTURUS: BUSINESS ANALYTICS CONSULTANTS


Creating Clarity from Chaos

Business
Intelligence

Enterprise Planning

Predictive Analytics

Our Team:

Business depth combined with technical expertise.


Former CFOs, CIOs, Controllers, Directors, BI Managers

Copyright 2014 Senturus, Inc. All Rights Reserved

700+ CLIENTS, 1500+ PROJECTS, 13+ YEARS

Copyright 2014 Senturus, Inc. All Rights Reserved

System Performance Tuning

System Performance Tuning


Tune the system settings and services to optimize speed
and efficiency
Areas to review:
tuning for peak usage times
hardware scaling
logging level

optimized FM models
optimized OS

Tuning Overview
Common terms:
Connection = a thread
Report Server = IBM Cognos Dispatcher

Report Service Process = report process (i.e.

BIBusTKServerMain)

A dispatcher will use all available threads on one report


service before using another
Primary focus for tuning should be on number of batch
report service and report service processes
Spread requests across multiple processes
More processes with fewer connections performs better

System performance tuning


Group Dispatchers
Group dispatchers into folders to apply settings across all

servers in a group

Specify the Number of Processes


Specify the number of process for the report and batch

report services
Specify the number of low and high affinity connections

System performance tuning


Peak Demand
Most organizations have a peak period of demand, typically

during normal business hours


Determine the start and end time to control the number of
concurrent jobs

System Performance Tuning Request Wait Times


You can specify how long a users request will stay in the queue

before timing out


You can also limit the maximum execution the report service to
prevent users from executing long report jobs

Examine request affinity


Affinity refers to whether a request is assigned to a
specific server or if load balancing can assign it to
another server

Requests are defined as having high, low, or absolute


affinity
Request affinity is not configurable
To optimize performance specify the number of
connections a dispatcher can open based on the affinity
of requests
Maximum threshold for affinity connections is 5
High and Low affinity connections should not
collectively exceed 5 per process

Summary Formulas for Initial Tuning


High affinity connections per batch report process = 1
unchangeable/not applicable to batch execution

Low affinity connections per batch report process = 2


Maximum batch report service processes = 1 * # of CPUs

High affinity connections per interactive process = 1


Low affinity connections per interactive process = 2
Maximum affinity connections per interactive process =
2 * # of CPUs

DEMO
IMPROVING PERFORMANCE

OF

SORTED REPORTS

Challenges & Goals


Issues with large reports containing sorting

Removing the sort improves performance


Requires a change in Cognos Configuration
Can dramatically improve the performance of sorted
reports

Copyright 2014 Senturus, Inc. All Rights Reserved.

Improving Sorted Report Performance


Demo
Purpose: In order to allow sorted reports to access more memory we must make a
change in Cognos Configuration and restart our server.
1.

To connect to the Training server, go to the Windows Start menu, select Run, type MSTSC, and
press Enter.

2.

Paste the Public DNS address that you are given by the instructor into the Computer name
field. Click Connect. Log in as modeler, Admin1234.

3.

Once the Remote Desktop starts, click the Start button in the remote session.

4.

Select All Programs, and navigate to the IBM Cognos 10 - 64 folder and click on IBM Cognos
Configuration. Click Yes to any messages.

5.

Under the Environment section is the Sort buffer size in MB. The default value is 4 MB.
Typically this can be safely increased 4 to 8 times. Increased sort buffer capacity will improve
ad-hoc query performance.

6.

Change this value to 32.

7.

On the toolbar, click Save. Click Close.

8.

From the Actions menu, click Restart. Click OK and then Continue to any warning messages
and then click Close.

Improving Sorted Report Performance


Demo

DEMO
CONFIGURING

DISPATCHER

IBM Cognos System Administration


Demo 8-1 System Tuning
Purpose: Lets look at how our system is currently performing
1.

In the Remote Desktop session, right-click the Taskbar, and select Start Task Manager.

2.

Click the Performance tab. Notice that there are 2 CPUs and about 7.5 MB of RAM. We want
to configure our processes to maximize the capacity of our dedicated server.

IBM Cognos System Administration


Demo 8-2 System Tuning
Purpose: Based on system utilization we can adjust some of our settings on the main
dispatcher to improve performance.
1.

Return to IBM Cognos Connection.

2.

From the Launch menu select IBM Cognos Administration.

3.

Click the Configuration tab.

4.

In the left side pane, click Dispatchers and Services.

5.

Next to the :9300 dispatcher name, click Set Properties and then click the Settings tab.

6.

Under Category select Tuning. Set the Entries range to 1 104. (yes, there are 104 Tuning
settings across all services)

7.

Because this system has multiple fast processors, change the Processing Capacity to 2.0.
In a two dispatcher environment, this dispatcher would handle 2/3 the processing while
the other would handle 1/3. This setting is relative to the total available Processing capacity
of the environment.

8.

Under Maximum number of processes for the batch report service during non-peak period
and Maximum number processes for the report service during non-peak period set the
value on both of these to 4. The best practice for these values is 2 per CPU.

IBM Cognos System Administration


Demo 8-2 System Tuning
9.

Under Number of high affinity connections for the batch report service during non-peak
period set the value to 4. Under Number of low affinity connections for the batch report
service during non-peak period set the value to 8.

10. Under Maximum number of processes for the batch report service during peak period and
Maximum number processes for the report service during peak period set the value on
both of these to 4.
11. Under Number of high affinity connections for the batch report service during peak period
set the value to 4. Under Number of low affinity connections for the batch report service
during peak period set the value to 8.
12. Under Number of high affinity connections for the report service during non-peak period
set the value to 4. Under Number of low affinity connections for the report service during
non-peak period set the value to 8.
13. Under Number of high affinity connections for the report service during peak period set
the value to 4. Under Number of low affinity connections for the report service during
peak period set the value to 8.
14. Click OK

IBM Cognos System Administration


Demo 8-2 System Tuning - Results

Scalability Horizontal vs Vertical

Scalability Planning
Scalability is the ability of a system to adapt to
increased processing demands in a predictable way
IBM Cognos was designed for scalability

It scales vertically using more powerful computers, and


horizontally using a greater number of computers
By default a weighted Round-robin load balancing
strategy is used when multiple dispatchers are available

Scalability planning Load Balancing


Use multiple instances of the components to balance
the load of requests
At the level of the dispatcher, weighted round-robin
load balancing is automatic in a distributed
environment
Web Server
Gateway

Server with
Dispatcher

Server with
Dispatcher

Server with
Dispatcher

Weighted Round -Robin Load Balancing


By default, each dispatcher has a relatively equal
weight, or Processing capacity, of 1
By assigning heavier weighting to a dispatcher, we
expect that dispatcher to process a larger volume of
requests in relation to its relative weight to the others
This Dispatcher will
receive twice as many
requests as the others.

Web Server
Gateway

Dispatcher
Weight = 1

Dispatcher
Weight = 1

Dispatcher
Weight = 2

Vertical Scaling
Installing multiple dispatchers on a single server can
leverage available capacity
You must install into a separate directory from the first
install
Remember to configure unique ports for each
component

Infrastructure Issues

Planning Infrastructure Components


Performance depends on the characteristics of your
infrastructure
Ideally, the IBM Cognos server components should be
connected by a network with 100 Mb of available
capacity
Network bandwidth between a Web browser and a Web
server does not affect system scalability, but does
affect user performance

Use true server computers, rather than fast


workstations

Planning Infrastructure Components


Install only gateway components on server computers
that are dedicated to Web server processing
Use the gateway type most appropriate for your
environment
The complexity of your security infrastructure can
increase response time

Because notification service generates additional email


traffic, ensure that your mail server can scale to support
the increased load

Other Topics

Advanced Topics
Optimized FM models
Use true dimensions and facts

Model from a data mart/data warehouse

Multi-tenancy
Supports multiple customers/organizations on a single

deployment
Helps restrict data to which users have access
Use the audit logs to keep track of tenants usage

Setting routing rules


Group dispatchers together

Assign packages and/or roles & groups to a server group


Allows certain heavy packages or users to be on a specific

dispatcher

Senturus, Inc. Proprietary and Confidential

Q&A

THANK YOU

You might also like