Are you attending Data and AI Forum in Miami next week? Want to talk some Cognos with me? Snag a half hour of my time to chat free from sales guys.
Thank you to those who attended Tuesday’s 11.1.4 livestream! Technical issues aside I had a great time doing it and will be hosting more in the future. There were some comments about performance in recent releases that I promised to address but didn’t – so here are my Cognos 11.1.4 performance tips!
Tip one: Dispatcher > Gateway
Cognos 11 brought with it the option to dispense with the gateway entirely and sending your traffic directly to the dispatcher. This is, among other things, a performance hack that can pay big dividends, particularly if your Cognos environment takes an extremely long time to load or if the UI feels generally sluggish.
The downside here is that for most enterprise deployments relying on a single dispatcher is just not an option for a ton of reasons – including little things like SSO, load balancing and disaster recovery. But if you are able to try this you might find it helps quite a bit.
Tip Two: Reality > Minimum Requirements
The minimum requirements for Cognos 11.1.4 are currently listed at two CPUs. This is a straight up no-go for using Cognos in almost all circumstances regardless of what version you’re on so long as that version starts with the number 11.
A tale of two configs
I typically deploy a single-instance Cognos machine on AWS with the following specs and have no performance issues for demos or small deployments:
- ec2 t3.xlarge
- 4 vCPUs @ 2.5Ghz
- 16GB RAM
With 11.1.4 this configuration suffers from a laggy UI and poor query performance, especially in Dashboards. I observed CPU utilization hitting 100% frequently. As a remedy I migrated my AMI to:
- ec2 c4.4xlarge
- 16 vCPUs @ 3.4Ghz
- 32GB RAM
This adjustment resulted in extremely snappy performance throughout the UI and like-for-like dashboard query execution improved from ~25 seconds to ~2 seconds.
Why did this jump so much?
Cognos 11.1.4 is doing A LOT more than 11.0.x or 10.x ever dreamed of with a host of new interactive services running in the background – services that do things like natural language query, automatic visualization or dashboard creation, on-the-fly predictive forecasting, integrated data prep and modeling…
Tip Three: SSD > HDD
One often unexplored area of performance improvement for Cognos 11.1.4 (or anyone using Transformer for that matter) is increasing the speed of your HDD or migrating to an SSD. All of the new features rely on either the Cognos 10 DQM engine or the Cognos 11.1 ‘Flint’ spark-sql engine to process queries. In either case data is written to and read from disk so IO becomes larger factor in determining performance than ever before, especially for any interactive task.
Tip Four: Your Browser Matters Now
Go ahead and conduct an experiment for me – pop into Dashboards on your local machine, build a bunch of visualizations and observer your resource use. You may notice that Cognos running in Chrome is capable of maxing the CPU on your laptop. It is capable of maxing the CPU on my monster gaming desktop as well if I push it hard enough.
So the final place I’d look to juice performance is actually at the hardware your users are toting around with them. It may seem a little unfair but just imagine life as a poor Power BI or Tableau administrator – outdated and slow laptops define the entirety of their user experience.
Cognos 11.1.4 Performance Summary
All of this is to say that I have had success increasing performance by significantly increasing the specs of the application server. This isn’t necessarily shocking but it’s comforting in a way because it means the problem isn’t bad code. It also means that if you’ve been skating by on the same server config you put into place in 2015 it’s probably time for an upgrade.