Cognos 10.2.2 Performance Tuning

Todd Schuman
Cognos Architect
Senturus, Inc

Q: Will PowerPlay Studio also be deprecated?
A: There are currently no plans by IBM to deprecate PowerPlay Studio.

Q: What are the licensing implications of distributed environment? Is a license required for each server, or does each user license it?
A: It depends on the licensing model you have in place with IBM.

Q: IBM recommends 4GB for the JVM, but you are referencing 1.5GB. Why?
A: This is according to IBM documentation. People have been hearing this from IBM Support for specific issues but it has not been documented at this time. More information on this topic.

Q: Does the ISAPI module work with Apache WebServer?
A: ISAPI is only available with Windows IIS.

Q: To confirm, Query Service does not need to be enabled if you are not using Dynamic Query Mode?
A: That is correct. The Query Service is for Dynamic Query Mode only and can be set to false to free up resources.

Q: How do you tune a report output that contains graphics?
A: We would need to review your issue before being able to provide a recommendation. You can contact us for a complimentary performance and stability review via email info@senturus.com, or by calling 888-601-6010, ext. 85.

Q: To run report in Dynamic Query Mode, do we have to convert all relational packages into DMR models?
A: That is correct.

Q: In a distributed environment, how many dispatchers can we add?
A: There is no limit to the number of dispatchers you can add.

Q: Can non-Dimensionally Modeled Relational(DMR) relational models go against JDBC data connections? if yes, would this boost performance over regular data connections?
A: Yes non-DMR models can go against JDBC connection. You would need a DMR and to define the package as Dynamic Query Mode (DQM)to take advantage of any performance enhancements.

Q: Can you explain more about how summary tables work?
A: Summary tables help performance when working with large amount of data that needs to be aggregated in real time; for example, if you have a table with 1 million sales records for each fiscal year. To get the annual total, you would need to select the >1 million rows and sum/group them up in real time. A summary table would have one line with the annual total for each year, so returning the amount would be a basic one row select statement.

Q: What are ‘stitched queries’ in Framework Manager?
A: Stitched queries occur in report when there are multiple fact tables being referenced; for example, a forecast fact and a sales fact. Selecting a measure from a conformed time dimension such as year will cause Cognos to determine the year and sales in one query, and year and forecast sales in another, and join them together. This can sometimes lead to poor performance depending on the amount of data that is “stitched” in Cognos.

Q: In Cognos on Linux, do you recommend CGI or mod2?
A: CGI is an older technology. We would recommend not using CGI if possible.

Q: Regarding migration from 10.2.1 with Tomcat to 10.2.2 with Websphere, what benefits or disadvantages?
A: So far, we have not seen any difference.

Q: Do you have a document that lists which services should be disabled when you have a dispatcher on the gateway host in a distributed environment?
A: We don’t have any specific document; we would need to review your existing environment.

Q: Are there any connection issues while using Vertica DB in Cognos that you are aware of?
A: Not that we’re aware of. Versions 4-7.1 should all be supported. More information.

Q: How do you tune presentation requests during peak periods when the volume of requests is very high?
A: We would need to review you environment to see where the root cause for performance is.

Q: Do you have any online training for System Administration?
A: Yes! See our training course schedule.

Q: How stable is 10.2.2 vs 10.2.1? I’ve seen that IBM doesn’t support SQL Server 2014 on 10.2.1, so we’re hesitant to upgrade our SQL Server instance. But I’m also worried about 10.2.2. Should we wait for 10.2.2 fix packs before going to 10.2.2?
A: So far, 10.2.2 has been very stable. FixPack1 was released mid-April, and there has not been any other fix packs since. 10.2.1 seems to have been a problem child of sorts as it is on FixPack7.

Q: What’s a good way to optimize Query Service (JVM) Memory in Dynamic Query Mode mode for reporting using a OLTP datasource?
A: Are you referring to Dynamic Cubes? Read a Dynamic Cube hardware sizing guide that can help you.

Q: What methods, tools, utilities, etc. could be used for a Linux Cognos server to help diagnose intermittent system/network/database issues that impact report runs?
A: The most common methods are top/Iotop, IPTraf, and nmon.

Q: To tweak your server to run more interactive reports (we have lots of unused memory), do you typically just increase the number of processes and leave the high and low affinity numbers as is?
A: You should review your audit logs and user activity to determine what the balance of high vs. low should be. Once you set this, you can bump up the number of processes and see how the server is responding.

Q: To run a report in Dynamic Query Mode, do we have to convert all relational packages into DMR models? Will just publishing a relational package with Dynamic Query Mode enabled work?
A: You must create a Dimensionally Modeled Relational (DMR) and publish as a Dynamic Query Mode(DQM) package. Not doing so will run reports in Compatible Query Mode (CQM).

Q: Primary content manager often fails over to secondary content manager. Does this mean we don’t have required hardware resources to support the traffic?
A: This is not normal. Cognos should only switch to failover server if primary content manager is not responding. You may need to add resources to the primary server or unload some of the work that is being done on this server to free up resources for your content manager.

Q: I’m hearing confusion between the terms Dynamic Query Mode and Dimensionally Modeled Relational here. Can you explain?
A: Dynamic Query Mode (DQM) is new with Cognos 10, and takes advantage of the 64-bit Dynamic Query Engine and in memory technology. Dimensionally Modeled Relational (DMR) is a method of structuring your framework models to act like an OLAP data source. To use DQM on relational models, you need to use a DMR.

Q: I am interestedin the Cube Designer / Dynamic Cube designer class and Cognos Query Studio.
A: For a full listing of our current training course schedule, visit: https://senturus.demo.techyscouts.dev/training

Q: We’ve experienced data retrieval issues that we believe are related to using local cache settings (framework model level setting). What is your suggestion on using local cache (related to performance vs. data integrity)? We have a six-dispatcher server configuration.
A: We would need to know more about your existing environment to know the root cause. Please contact us for a complimentary performance and stability review via email info@senturus.com, or by calling 888-601-6010, ext. 85.

Q: Have you experienced CPU spikes due to Garbage Collection when the JVM was running out of memory due to high concurrency Dynamic Query Mode reporting?
A: We have seen this, but diagnosing the issue can be tricky and requires some real-time monitoring. Contact us and we can help. You can contact us via email info@senturus.com, or by calling 888-601-6010, ext. 85.

Q: What are you doing for clients who must report against OLTP?
A: There are many things that should and should not be done when working with OLTP. Please contact us for more information via email info@senturus.com, or by calling 888-601-6010, ext. 85.

Q: Do you need to configure Cognos any differently for summary tables? Is Cognos Framework Manager aware of summary tables?
A: A table is a table in the eyes of Cognos Framework Manager. The benefit is that the number of rows and amount of aggregation will be reduced tremendously when using a summary table.

Q: We get 100% CPU usage on our dedicated PowerPlay virtual server with just two cubes building. The server has two Intel Xeon E5-2650 processors and 4GB of memory but memory usage is low. Does PowerPlay not leverage multiple cores/processors?
A: Transformer cube builds can be resource heavy at times. To determine the correct sizing for building multiple cubes, we would need to review the cube models and amount of data. Contact us via email info@senturus.com, or by calling 888-601-6010, ext. 85.

Q: If I need to decide between Active Report and Workspace Advanced, which is a better approach if my volume of data is a concern?
A: Active Report will pull down data and store it within a .MHT file. Once an Active Report file has been created, no database connection is needed. This may be a good solution if you will have many users accessing the same information. If your users require more ad-hoc reporting that cannot be addressed with a single central report, it may make more sense to use Workspace Advanced.

Q: I’m having an issue logging in using SSO. I logged in as a different admin account on my machine, and now it logs me in as that administrator all the time. How do I clear the information?
A: We would need to review your current configuration to troubleshoot.

Q: Are there any webinars on browser compatibility issues using Cognos 10.2.1 / 10.2.2?
A: The full list of supported browsers.

Q: Active Report sometimes doesnot open in 10.2.1 workspace. The server becomes really slow. Are you aware of something like this?
A: I am not aware of any issues. We would need to review your reports and environment to troubleshoot.

Q: Can core.dmp files be deleted periodically? I see lot of them occupying huge space on the server example.
A: Core dump files are usually an indicator of a serious problem. You can delete them, but steps should be taken to prevent them from being created.

Q: What does pogo failure mean while starting your secondary content?
A: This could be a result of many things. We would need more details to troubleshoot. You can contact us for a complimentary performance and stability review via email info@senturus.com, or by calling 888-601-6010, ext. 85.

Q: How much additional resources are needed for Active Report?
A: I am not aware of any additional resources needed for Active Report. There are steps for improving speed and keeping file size down that can be found on the IBM Knowledge base.

Q: We just changed from 10.1.1 to 10.2.1. Should we change to 10.2.2, and is easier to do so?
A: Upgrade process is the same. There are many less fix packs to apply in 10.2.2.

Q: Can we have different summary tables for different grains, like hourly, daily, monthly tables, and have common time hierarchy?
A: Yes, although it will require more modeling and work to maintain. I have often seen views used to union many different summary grains. A dynamic or power cube would be able to do this for you automatically based on your dimensions.

Q: In Framework Manager, what is the use of semi-aggregate?
A: Semi-aggregate is for SAPDW and should not be used for other data sources.

Hardware Performance

• Sizing Up Servers: Processors, Sockets, and Cores
  • Processes
    • An independent program running on a computer. A process has a full stack of memory associated for its own use, and does not depend on another process for execution.
  • Thread
    • A thread is essentially a process that does not have a full stack of memory associated for it. The thread is tied to a parent process, and is merely an offshoot of execution. Typically thread processes must run on the same computer, but can execute simultaneously on separate cores of the same node.
  • Hyper-threading
    • Hyper-threading is an Intel technology that originally preceded multi-core systems, and was used to make a single core appear logically as multiple cores on the same chip. Intel abandoned hyper-threading briefly during the advent of multi-core processors but reintroduced the technology in 2008. Since then, Intel has used it extensively to improve the performance of parallel computations in its multi-core processors. Hyper-threading improves performance by sharing the computational workload between multiple cores whenever possible, allowing the operating system to schedule more than one process at a time.
• Adjusting RAM
  • Use the following memory settings as a starting point and adjust them based on the memory usage of your system.
    • 2 GB for the base operating system and accompanying software, such as antivirus, back up, and enterprise management software
    • 4 GB for a 64-bit Content Manager JVM
    • 4 GB for a 64-bit Application Tier JVM
    • 2 GB for the graphics JVM (IBM Cognos Workspace)
    • 2-4 GB for the query service (dynamic query mode) JVM
    • 1 GB per core for the report services processes (dynamic query mode) (JVM)
    • 2 GB per core for the report services processes (compatible query mode) (BIBuS)
  • Note: Dual Core, 64-bit Single Server needs 16GB for CQM
• Distributed Environments
  • Load balancing
  • Fail over
  • Hardware aligned for specific Cognos needs
  • Scalable
• Virtual vs. Physical Servers
  • Pros
    • Reduced maintenance on physical servers
    • Easily allocate additional resources
  • Cons
    • Overhead: 5-10%
    • Dedicated resources

Cognos Server Tuning

• Concurrent Users
  • 100:10:1 Rule
  • 100 Named Users
  • 10 Active Users
  • 1 Concurrent User
  • 4,000 Named Users = 40 Concurrent Users
• Dispatcher Tuning
  • 93 Tuning Categories!
  • Batch Report Processes
  • 1 * # CPU Cores, 2GB CQM, 1GB DQM
  • Interactive Report Process
  • 2 * # CPU Cores, 2GB CQM, 1GB DQM
• Dispatcher Requests
  • High Affinity Tasks
    • Report Viewer links (Run again, Return)
    • HTML report navigation(Top page, Page up, Page down, Bottom page
    • Delivery options(Save, Save As, Print, Email, Viewing)
  • Low Affinity Tasks
    • Report Querying(Reporting, Report Processing)
    • Report Authoring(Metadata Retrieval, Query Validation)
    • Administrative(Testing Data Source Connections, Adding Objects (Folders, Jobs, Schedules, etc.), Refreshing Portal Page)
  • Peak/Non-Peak Hours
• WebSphere Liberty Profile
  • To reduce the startup time, memory footprint, and resources used, use the default setting of 768.
  • To balance between fast startup time and quick operating speeds, type a value about 1.5 times the default value, such as 1152.
  • To maximize operating speeds and if performance is more important than fast startup time, and if your computer has a lot of resources, type a value about double the default value, such as 1536.
• Quick Fixes
  • ISAPI instead of CGI
    • Improved management of multiple users
  • IIS – Content Expiration
    • COG_ROOT>/webcontent/pat/images
    • 600+ Images
    • Set HTTP Response Headers
    • Common Headers
    • Expire Web Content
  • Dynamic Query Mode (DQM)
    • Not using DMQ? Turn off Query Service to gain 1GB RAM each server

• Performance and Scalability Testing
  • End-User Transactions and Load Generation Tools
    • IBM Rational Performance Tester
    • HP Loadrunner
    • Telerik Test Studio (previously Fiddler)
  • Server Resource Tools
    • Microsoft’s SysInternals Tools (Windows)
    • Microsoft’s Performance Monitor (Windows)
    • NMon (AIX and Linux)
    • ‘top’ and ‘ps’ (Linux and UNIX)

64-bit and Data / Report Tuning

• 64-bit Advantages
  • Access more memory
  • 64-bit registers allow faster, more efficient execution
  • Stability
• DQM / Dynamic Cubes
  • In-memory Aggregation
  • Relational DMR
• Run DQM
  • Install JDBC Drivers
  • Enable JDBC Connections
  • Enable JDBC in Framework Manager
• Senturus.com/Resources
  • Beyond PowerPlay: Choosing OLAP
  • Dynamic Cubes in 10.2
  • IBM Cognos Dynamic Cube Deep Dive
• Star Schema
  • Less join paths
  • Allows quick understanding, navigation, and analysis of large, multidimensional data sets
  • Flexibility
• Summary Tables
  • Pros
    • Eliminate need for large, on-the-fly aggregation of large tables
    • Fast performance
  • Cons
    • Requires DBA / maintenance
    • Built at only one level of aggregation
• Power Cubes
  • Pros
    • Drill up / down
    • Aggregation at every dimension
    • Cognos dimensional functions
  •  Cons
    • File size
    • Latency
    • Build time
• Other Performance Causes
  • Reporting Design
  • Database Performance
  • Network / Firewall
  • Stitched Queries
  • Local Processing
  • Infinite Others