en
nl
Accelerate your dashboard with performance evaluation
Qlik Cloud
Within the Qlik Cloud platform, you know everything about how your dashboard is performing and what you need to do to improve it at the touch of a button. For each app, you can have a performance assessment performed and see the response times for each visualisation or sheet. It is also possible to compare different performance assessments and assess whether work has really improved. When users complain about a slow dashboard, the performance assessment guides you to the solution.
Perform performance review
The performance review is started from the HUB, by clicking on the 3 dots behind the app and choosing 'Evaluate performance' or from the app details under Performance review. Then click on 'Evaluate now'. After the evaluation has been performed, you will immediately get an insight into the results.
First of all, these are the app's general data, covering the size of the app, the number of rows of data and the number of published worksheets and objects. Note that only published sheets and visualisations are assessed. Other users' own worksheets, for example, do not influence the results in this way.
In the results tab, you will find a number of categories where long load times occur. When you click these open, direct references to the objects are available.
- Objects with cache problems:
These are objects that are not cached efficiently. This may be improved by modifying expressions or the data model.
- Objects with one thread:
When objects appear here with excessively long load times, bottlenecks in the calculation should be sought. These occur with, for example, virtual tables (Aggr() function) or the use of multiple fields from different tables in a complex data model. - Objects exceeding memory limit:
Occurs when complex calculations exceed the licence limit
- Initial load time per worksheet/Cached load time per worksheet:
These are measured by opening each worksheet 2 times. The first time the initial load time is measured and the worksheet is cached right away. The second time the worksheet is opened it is cached and should load faster. Object loading time is also considered here for each worksheet.
Furthermore, in the results we also find the top 5 objects, tables and fields with high load time or high memory allocation.
Assessing the results
Now that we know what the assessment results are, we need to start doing something to improve performance. We know which objects have a long load time, and which fields and tables in the model get a high memory allocation.
Changes to the data model often lead to the biggest improvement in an app's performance. For example, consider simplifying a snowflake to a star model, as every 'hop' between 2 associated tables affects performance.It also makes a significant difference to replace key fields with an Autonumber(). Comparisons over numeric values are faster than comparisons over (long) string values. Also, when using an AutoID, the result is equivalent to the index table in which Qlik stores unique values, which is even more profitable!
Also, it is a good idea to limit fields with long and many unique (string) values. These are fields that you can find in the 'Top 5 fields with the highest memory allocation' list. Assess whether these fields are really necessary for the analysis (e.g. description and free text fields). If not, remove them from the data model.
Request
Try Qlik Cloud free for 30 days now
Try our Qlik Cloud for free now! Sign up and after 30 days the free trial period ends without any commitment. Completely non-binding!
AanvragenShortening long unique values
In a number of cases, long unique values can also be shortened. For example, consider fields with a TimeStamp in which only the date is important. A Floor() function is then used to round off the timestamp. Even if the timestamp registers to the second, but only the hour level is important for the analysis, gains are made here.
Besides changes to the data model, we also optimise the calculations used on the sheets for performance. The calculations often seen as the most 'heavy' are those where if() or aggr() functions are used.There are several efficient alternatives to features that can reduce loading time. Such as:
The if() function
When the if() function is used within an aggregation, it is executed at the line-level. That means the evaluation takes place for each data line from the aggregation. Thus, the function could be executed thousands of times. If this occurs, the function is replaced by set analysis. Also, all possible outcomes of the if() function are calculated, not just the expression valid for the outcome being displayed. If possible, avoid the if() function in worksheets and use set analysis.
The Aggr() function
The Aggr() function is a very powerful function used, among other things, to perform nested aggregations. However, it is possible for this function to underperform a dashboard quite a bit. In these cases, see if the calculation can also be made in the script, using, for example, a TOTAL qualifier, or a weighted average by division without Aggr.
The Count(Distinct...) function
The same applies to calculations using Count(Distinct ...). Normally, calculations are performed multi-threaded, however, a count distinct requires all unique values, so it too is performed single-threaded, unless the fields used for dimensions and measurement come from the same source table in the data model.
Contact us now!
As you have just read, there are a lot of factors that make a dashboard slow, but fortunately also a lot of solutions that help you with this. In the unlikely event that you still cannot figure it out, contact us and we will take a closer look at your Qlik dashboard!
