Author Topic: Cognos scaling  (Read 4820 times)

Offline bartburg

  • Full Member
  • ***
  • Join Date: Jan 2014
  • Posts: 38
  • Forum Citizenship: +0/-0
Cognos scaling
« on: 10 Apr 2014 03:25:28 am »
I dont know if the topic is the right one correct me if it is not!

My questions:
- we have a organisation that has 1000 users
- 28 database connections in cognos
- we run around 1900 reports per week
There are more application managers that want to connect to cognos every day so that means more connections and more reports
how do i Ensure the quality of the server?
server stats
VMWare server 2008r2 64bit
2 processors @ 2.33GHz
8GB memory
IBM cognos10

i hope some of u guys have experience with this sort of growth

Offline bdbits

  • Super Moderator
  • Statesman
  • ******
  • Join Date: Feb 2010
  • Posts: 1,762
  • Forum Citizenship: +104/-0
Re: Cognos scaling
« Reply #1 on: 14 Apr 2014 10:18:55 am »
These are just some ideas for things you might want to consider.

This is capacity planning, no different because it is Cognos. First thing is to decide what to measure - CPU utilization, disk I/O, RAM utilization, etc. Then baseline your current system. It helps if your organization has some monitoring tools here, as you probably want to sample it over time to get averages. Then rerun your measurements at regular intervals, maybe make a graph showing the changes over time. Based off of this you can guess at your future requirements and plan to adjust accordingly.

1900 reports a week over a 5-day 40-hour workweek would be a little less than one report a minute on average. Whether this is a lot or a little obviously depends on the particulars of your data and hardware, which affect how long each report takes and resources used, as well as whether the report runs are clustered around particular times of the day or month. If your reports are predictable and repeatable for multiple users, you might look at running some at off-hours on a schedule and saving the output version, or bursting them. There are things you can do database-wise to speed things up, or use cubes if appropriate. There are too many variables to give very specific advice, it depends on your situation.

Offline erikb10

  • Associate
  • **
  • Join Date: Mar 2015
  • Posts: 3
  • Forum Citizenship: +1/-0
Re: Cognos scaling
« Reply #2 on: 24 Mar 2015 08:28:42 am »
I know I am coming late to this discussion.  But there are some important things that need to be pointed out.

You have to remember that Cognos is a J2EE application.  And that means that it runs completely inside the Java Application Server you have installed it with.  For most of us that will be Tomcat.  And your Java Application Server runs completely inside a Java Virtual Machine.  So if you have not realized it before, Cognos is a sandbox application.  And it will never have access to, or use more ram than the ram stack that the JVM started with when the Java Application server was started that Cognos runs on.  I have found that the most installs have this set at somewhere between 1 and 4 GB of ram. 

There is a good reason why Cognos was designed to be installed as a multiple machine distributed application. 

Your mileage may vary, but I have found that somewhere between 100 and 200 users is the point where you want to have your second application server set up and running.  This experience sort of matches IBM's 100:10:1 rule about Cognos capacity planning.   

So with 1000 users, you probably want between 5 and 10 application servers. 

As for how many reports you run in one day, it is less about that and more about how many you run in any given hour and the average run time for each in that hour. 

Try writing a report against your audit database that tells you how many reports run each hour of the day, and what the average run time is for each hour.  you will find that information far more informative of load throughout the day than the overall daily count. 

I hope by now you already have a solution.  But I also hope this helps with your future planning. 

 


       
Twittear