Skip to main content
Announcements
Have questions about Qlik Connect? Join us live on April 10th, at 11 AM ET: SIGN UP NOW
cancel
Showing results for 
Search instead for 
Did you mean: 
Tyler_Waterfall
Employee
Employee

Governance Dashboard: Why do you use other system tools (as well)?

I am seeking your feedback on on the QlikView Governance Dashboard 1.0 GA available on QlikMarket and on the QlikView download site.

As you may or may not know, the Governance Dashboard provides visibility into your QlikView deployment, showing operational statistics for server and publisher and also extensive metadata about QVWs, QVDs, QVXs including data lineage, sheet objects, expressions, etc.

For those currently using the Dashboard (or who have deliberately chosen not to use the Dashboard), I am interested in knowing why you use other tools (e.g. Ops Monitor, MetaScanner, or a home grown/other tool(s)) to monitor and manage your QlikView deployment.  Any feedback and questions would be great! I will use this feedback in prioritizing and, where needed, augmenting open requests for future releases of the Dashboard.

Thank you! Gracias! Obrigado!

Tyler

14 Replies
Tyler_Waterfall
Employee
Employee
Author

Rob,

Thanks for sharing.

It looks like you are creating a concurrent session count by second and then aggregating that up to the hour, correct?

I like the approach, but my challenge is to create an approximate concurrency count (for guidance on licensing) where I could see thousands of sessions spanning months and even years. How do I manage that analysis without blowing up the data?

One approach I've seen creates session counts by hour which could still lead to a lot of data, but much reduced.

I'm not sure what my strategy will be at this point.

Tyler

rwunderlich
Partner Ambassador/MVP
Partner Ambassador/MVP

Tyler,

Yes, I create a concurrent count by second. Then I save the max session count per hour. I've used the data for licensing recommendation and monitoring. You have to aggregate the max at some level -- hour, day, whatever.

The script processes only the last two days, so at most interval matching 86,400 * 2 seconds -- no big deal. I run it that way at customers and collect the ongoing data in a QVD. When processing a customer's logs for the first time I've bumped that count up to 14 days, still runs quick -- that's only 1.2M second values. I think you could do a month easily even with thousands of sessions. You could do it in weekly chunks as well.

I've seen the session counts by hour method in other dashboards, where a session is credited in the hour it starts or in any hour it exists. I don't think that's accurate enough for concurrence studies. My technique computes the same value -- max concurrent sessions per hour -- but more accurately.

-Rob

Tyler_Waterfall
Employee
Employee
Author

Stu,

Thank you for your response.

Your 'external' license monitor app sounds familiar, unfortunately.  I can't promise anything specific for the Governance Dashbaord, but I will say that I am working to get some visibility into license usage and concurrent sessions by Cal Type, which will help in your analysis, though it will not provide the complete picture you describe.

Would you mind sharing the chart that you described which lets you

know how often users are accessing it and, critically, how many clicks they are taking to find the information they want.

It can be a screen shot or even just the list of the dimension and expressions. I'm gathering ideas in this area.

Thanks again.

Tyler

Tyler_Waterfall
Employee
Employee
Author

Rob,

I think you will have more 'intermediate' rows than 86,400 per day as you have it grouped by Server and Cal Type.

Regardless, though, the number of rows gets very large very quickly as you add more days.

I like the 'incremental' approach, though it requires an initial load that might itself be tricky to implement.

In your estimation, how accurate would an "hour window" approach to concurrent sessions be? Or, stateed differently, if your option were to have this "hour window" concurrent session data or no concurrent session data, which would you prefer? (Sometimes 'almost correct' data is worse than no data...)

Tyler

rwunderlich
Partner Ambassador/MVP
Partner Ambassador/MVP

In my experience, the "hour window" method is not accurate enough. In a gross statistical sense, it's pretty close. But in any given scenario, it can be widely inaccurate, and tends to overstate the actual concurrency. That's crucial if you are telling a customer "based on these counts, I recommend you buy N session CALs". If anyone audits by hand  a particular time period , they will find the numbers to be incorrect. 

-Rob