Installing Mulitple Disptachers on a Single Server

[jstouffer]

Does anyone know if this is possible and if so where I can find the documentation on how to do it?  I've searched the KB but have not found anything specific.  Only vague references.

[ducthcogtechie]

Just a step back here.
Why do yo want it, and what are your server specifications?

[jue]

Yeah, it's certainly possible. We have a single Cognos environment running 20 separate Cognos component installs across 10 UNIX servers.

As 'dutchcogtechie' explains, there are potential pitfalls to running a distributed environment and you have to be clear why you want to do it. The main reasons for doing it are scalability, resilience and performance, performance, performance!! But there is a cost in terms of increased maintenance, troubleshooting, complexity, upgrade issues etc. And more often than not, Cognos releases are not tested on distributed environments so you tend to come across more inter-component communication problems than single server installs.

It's not complicated to set up at all but perhaps more than can be written in a forum reply !!!

[goose]

Essentially all that is required is to run different port no.s for the relative components between the installations. However as stated in previous replies what is your motivation for doing so? 

Personally I have only run multiple instances on the same box for development and testing of reports etc over multiple Cognos releases.

[jstouffer]

I am trying to help out mu IS department.  We have been having performance issues for some of our more complex reports.  We are reaching the 2G limit for the Java instances and from what I understand creating a new dispatcher along with the reporting services will help to maximize our RAM usuage.

Am I correct that basically we just need to to a second install of C8 on the server and assign new ports to the services?

[ducthcogtechie]

An additional dipatcher will not make your current reports a second faster.
Report calculation is done in the bibus proces, wich has it's own 2 gig limitation.

i am trying to help here; what are your server specifications, user load etc?

[jue]

It depends on whether your report is slow because of high usage on the environment (i.e. lots of users) or whether the report is running a large/complex query.

Run the report when when the environment doesn't have a high user load. If it runs slow, then it will still run slow on a multiple instance enviroment. Nothing you can do other than redesign the report or optimise the database you are reporting from.

[goose]

We have been having performance issues for some of our more complex reports.

complex reports = crap sql/mdx = slow response 99.999% of the time  If you are having performance issues for some complex reports start there by making sure they are optimized or by making them uncomplex. Users don't always "have" to have the data the way they "want" and if they are waiting for a few minutes for the report to return in a few months the reports wont even get used anymore.

For Reports on Relation Data Sources

1. Trap SQL generated by reports, preferably using same parameters users have when they complained about performance

2. Make sure all joins/filters are valid/there and generate an explain plan, or get your DBA to do some QA on the SQL

3. Make sure all report filters getting sent through the db, sometimes depending on how you filter Cognos can only apply filters in its own app engine which means if you have a big fact tables you could be getting millions of records back when you only want a few thousand.

4. Don't be afraid to bypass FM, sometimes its the only way to drastically improve SQL response times. Sure we all want to use FM but hey in reality we aint got GO Sales for a data warehouse ;D There some SQL concepts that cannot be modeled in FM

5. Capitalize on new features in Cognos 8.2+ that allow queries to be run in parallel under certain conditions

6. I have reduced report run times ten fold by simply changing compound index column orders and changing filter mechanism... it can be that simple

For Reports on Mutli-Dimensional Data Sources

1. Make sure you are excluding null values for cell calculations as OLAP queries cross join member sets from different hierarchies

2. Avoid attribute filtering if you can

3. Read http://support.cognos.com/supported/tti/public/docs/cognos_pp_reporting_writing_efficient_olap_queries.pdf?lid=cognos_pp_reporting_writing_efficient_olap_queries


Good luck by the sounds of it your issues are not related to capacity but a few slow reports.