Cognos Analytics Performance Tuning

Todd Schuman
Practice Lead – Installations, Upgrades and Performance Tuning
Senturus, Inc.

Todd heads up the Installation, Upgrade and Performance Tuning practice at Senturus, bringing more than 16 years of Cognos and Tableau experience to his role. Since his early days in BI working with Cognos Enterprise Planning and Cognos Impromptu to today, Todd has worked extensively on developing and designing enterprise systems and reports. In addition to managing and performing multiple architecture engagements at Senturus, Todd works closely with the IBM Cognos product development team to stay abreast the latest enhancements and changes.

Q: Where can I get this presentation?
A: You can download this deck from our website at //senturus.com/events/cognos-analytics-performance-tuning/.

Q: How do we dedicate CPUs to certain pieces if we have our CM and dispatcher on the same server in Cognos Analytics?
A: There is not a way to “dedicate” a CPU to a specific process. The IBM recommendations are just a starting point for sizing up hardware. You should actively monitor the servers during peak use to see if you have more or less than you actually need.

Q: If I run “all” in single server what is the required RAM when using Cognos Analytics? In addition, what are the RAM specifications for dispatcher modules?
A: It depends on how many concurrent users you are expecting as well as other factors such as CQM vs. DQM and report complexity. We don’t recommend running with less than 16GB, but would prefer 32GB.

Q: My Cognos Analytics distribution system is well-over specified, but it is still slow. Is there any documentation or guidance on checking network traffic, content store speed and data source speed? I am trying to find the bottlenecks.
A: We are not aware of any specific document on next steps. Check the database first as that is typically the bottleneck in most performance related situations.

Q: How do I determine the number of cores using Cognos Analytics since most of my customers use the standalone installation?
A: You won’t be able to determine what cores are being used by Cognos Analytics, but you can check on the hardware by running MSINFO32 on a Windows server.

Q: Is there a way to let consumers see their report run activity in Cognos Analytics?
A: This was added to 11.0.8. Unfortunately, it’s not possible without Admin/Manage access prior.

Q: Can you describe where the IPA in Cognos Analytics is?
A: Under run options there is a menu screen. Click the last checkbox, which will allow you to enable this for a specific report.

Q: If I have Cognos 11.0.6 and 11.0.8 is available, can I install over the top of 11.0.6 (like service packs in Cognos BI 10) or is a fresh install required?
A: Yes, you can install over the top. Make sure to back up your content store database prior to the install. If you have any custom files, images or JS you will also want to make sure your preserve file is setup to protect them.

Q: What are the tuning recommendations for heap/garbage collection in Cognos Analytics?
A: This is an advanced configuration that would need to be reviewed on a case-by-case basis. We can help if you’d like a detailed review, contact us at 888 601 6010 ext. 87.

Q: Do data sets need extra memory in Cognos Analytics? We want to be prepared in our planning for the right amount of memory.
A: Data sets are Apache Parquet files that are stored in the content store database. They are brought down locally on the server when needed and should not require additional memory.

Q: Does Cognos Analytics support Performance Monitoring Analysis PMA (zippi database)?
A: We have not attempted to try this in v11. It was a bit experimental even in Cognos BI 10, so it would be up to you to use in CA11.

Q: Is it possible to delete or archive content administrative tasks under Administration in Cognos Analytics?
A: You can delete content management tasks such as imports, exports and upgrade tasks.

Q: Do you think performances are better on cloud IBM or on-premise servers?
A: There are many factors to take into account when answering this question. Cloud provides different levels of sizing from small to large with bare metal servers. On the other hand, you will have more control over your own on-premise hardware and configuration so the sky is the limit. It would be up to you to find the best fit for price vs. performance.

Q: What migration method do you suggest we use with full deployment or migration over the existing content store database in Cognos Analytics?
A: We always recommend using a content store database backup/restore.

Q: Is it possible to reduce scheduling capability in Cognos Analytics by assigning reports per day/hour by employee?
A: You can grant access to specific Cognos capabilities around scheduling. The options available are day, hour, minute, month, trigger, week, year and priority.

Q: Is there a way to restrict the search to prevent users from seeing content that they do not have access to in Cognos Analytics?
A: We are aware of this defect in prior releases of v11 and so is IBM, who is working to resolve this.

Q: What is the magic number for total number of affity per process in Cognos Analytics?
A: There isn’t a magic number, it’s unique for each environment. Please contact us at 888 601 6010 ext. 87 if you would like to review your environment.

Q: What are the hardware requirements for Cognos Analytics?
A: Please refer to the following link for IBM hardware recommendations: https://www.ibm.com/software/reports/compatibility/clarity-reports/report/html/hardwareReqsForProduct?deliverableId=2B4A13308DA311E7A1A213628837956C&
osPlatforms=AIX&osPlatforms=Linux&osPlatforms=Mobile%20OS&
osPlatforms=Solaris&osPlatforms=Windows.

Q: How does the hours setting work on Cognos Analytics? Most of my batch reports run from 7am to 11am.
A: There are many ways to deal with batch reporting, but if you want to run batches during non-peak hours, you could set peak start hour to be 11 and peak end hour to be 7. This gives priority to batch services during non-peak, and priority to report services during peak.

Q: Is there a recommended maximum number of recipients for burst reports in Cognos Analytics? We have a burst report that emails anywhere between 1,000-10,000 recipients. This report sometimes takes three hours to complete.
A: That sounds like a large burst, which would take some time. There may be some options to tune this process, so please contact us at 888 601 6010 ext. 87 if you would like to schedule a time to review.

Q: Is there a way to set default format for running a report in the background on Cognos Analytics?
A: We are not aware of any way to define a default format for only background related reports. However, you could try editing some of the source code, but this change would be unsupported and at risk for future releases.

Q: Can we use the search option to search by folder in Cognos Analytics?
A: We are not aware of any way to do this currently. You may want to request this functionality from IBM by using the RFE website:
https://www.ibm.com/developerworks/rfe/?BRAND%20ID=301.

Q: What is the total execution time available for all users with Cognos Analytics?
A: This defaults to zero, which has no limit. You can set it to something if you see a lot of long running reports that are eating up all of your resources.

Q: How do we run report activity for consumers in Cognos Analytics Release 8?
A: You can access the “My Schedules and Activity” by clicking on the small user icon in the top right corner of CA11.

Q: To increase performance in Cognos Analytics, do you recommend we save upload files in Content Store, in file system or in object store?
A: We typically stay away from using an object store as there are some risks involved with that approach. Click on the link to get additional information on where to store your uploaded files: https://www.ibm.com/communities/analytics/cognos-analytics-blog/where-is-data-uploaded-into-cognos-analytics-stored/.

Q: Is the total execution time (IPA) option available for all users?
A: This is for report developers. If you have the ability to edit reports, you can access this functionality.

Q: Can we choose size limit by roles/groups per upload of data in the configuration menu? The default is 100 MB, is that for all users?
A: Not currently, but if you would like to see it added, please request it via the RFE website at https://www.ibm.com/developerworks/rfe/?BRAND%20ID=301.

Q: Do you have any courses which cover those 90 different Cognos Analytics tuning options we could purchase? We’re very interested in learning to tune ourselves.
A: We cover a good many of the tuning options in our Cognos Analytics System Administration class. Or we can do a one day mentoring session.

Hardware and Server Specifics

• Use the following processor settings as a starting point and adjust them based on the use of your system
  • Content manager – 4 CPU
  • Dispatcher – 8 CPU
  • Server can have multiple sockets
  • Each socket can house a CPU
  • CPU can have 1-48 cores
  • Logical processors <> cores
  • Hyper threading
• RAM
  • Use the following memory settings as a starting point and adjust them based on the memory use of your system
    • 4 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
    • 1 GB for the Cognos graphics service JVM
    • 8 GB for the query service (Dynamic Query Mode) JVM
    • 1.5 GB for each batch/interactive report service

Architecture – Distributed Environments

• Basic distributed environment
  • Dedicated:
    • Gateway
    • Dispatchers
    • Content manager
• Distributed environment with failover
  • Active/standby content manager
    • Both pointed to same content store
    • First CM to start becomes active
    • Standby always checking for active up status
    • Will become active when primary becomes unresponsive
• High availability (HA)
  • No single point of failure
  • At least 2 of each Cognos component (gateway, dispatcher, CM)
  • Load balancer for gateways

Tuning

• High affinity activities
  • 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 activities
  • 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
• Cognos services
  • Running reports
    • Batch – running in the background
      • Jobs
      • Schedules
      • Emails
      • Saved output
      • Triggers
  • Report – interactive, user actively involved and watching screen
• Peak vs non-peak
  • 24 hour clock
  • Defaults
    • 7 (am) start
    • 18 (6 pm) end
  • Recommended
    • Give report service priority during peak
    • Give batch server priority during non peak
• Concurrent users
  • 100:10:1 rule
  • 100 named users
  • 10 active Users
  • 1 concurrent user
  • 4000 named users = 40 concurrent users
• Putting it all together
  • Affinity + Cognos service + peak/non peak
  • Be aware
    • Each connection is multiplied by the number of processes
      • 16 low affinity (8*2)
      • 4 high affinity (2*2)
      • 20 total connections
    • # of CPU * (2-4) = # of connections
    • 5 GB RAM * # of connections
• Report status
• Server monitoring

Virtual Servers

• Traditional vs virtual
• VM recommendations
  • Dedicated resources
  • Enterprise level host hardware
  • 5-10% overhead

Report Specific Tuning

• Data Explosion
  • Data
    • Queries/SQL
    • SQL execution plan
    • Joins/modeling
    • Star Schema/data warehouse
    • Indexes
    • Summary tables
    • Cubes (Transformer PowerCubes/Dynamic Cubes)
• Interactive Performance Assistance (IPA)
  • New with R7
  • HTML only
  • Enable
    • Run option dropdown
    • Show run options
    • Include performance details
  • Output
    • Execution time (in milliseconds)
    • Rendering time (hover)
    • Query time (hover)