Application Design

Introduction

Applications can denote both the applications for the end-users. Most notable are AppOne and the Connectors used for the data transformation by the Superadmins. Although the use cases are different, there are overlapping points to keep in mind while working in these applications.

Performance can be highly dependent on the specific application. When designing your application, you should always be aware of the impact on the performance of specific decisions.

AppOne is the primary example of an end-user application. It is delivered out of the box to customers and can be used as a basis for your own application. During developing apps and discovery accelerators, much attention is given to ensure optimal performance for the end-user. However, there may be specific details of your dataset and requirements which may require application optimization techniques.

Functional Tasks

The UiPath Process Mining platform has a lot of functionality with its own performance characteristics. For example:

Calculating expressions
Joining tables
Calculating filters
Process mining
Calculating trees
Data loading.

Most of these tasks depend only on the number of records in the table on which the operation is calculated. Data loading also depends on the number of columns. Some operations are faster when there is a low number of unique values in the input columns.

Analysis

Use the UiPath Process Mining Profiler for an application profiling run to get insight in loading times of your application and dashboards.

Solutions

Development

Dashboard items

Avoid too much information on one dashboard, especially too much detail, since these are all unique records that need to be calculated.
The amount of data on which dashboard items are based affects the performance. Dashboard items based on more data may be slower.
The number of items on a dashboard affects the performance. Dashboards with more items are slower. Change the Availability of table items to Private or Hidden whenever possible.

Filters/selectors

Set from/to root filter to TRUE and remove the end-user attributes for the from/to activity.
Set attribute selectors default value to NULL, where possible.
Disable grey values of filters.
Define the default value of a selector based on your dataset. The default value should not have too many unique values.

Graphs

Sort graphs on size.
Use the Advanced Process Graph settings to optimize the loading time of your process graphs. For example, use Simple hints instead of Extensive hints.

Expressions

Prevent user-dependent expressions.
Prevent filter-dependent expressions.
Complex expressions lead to higher response times. Keep expressions as simple as possible. Split up expressions where possible.
Expressions containing the self command are often a lot slower. It is recommended to use the prev command.

Overlays

Text overlays are calculated on a certain level, which determines how often they are calculated.
Overlays on records level are calculated many times, on root level only once.

Tables

Live tables are loaded when the user logs in and/or changes a filter control. Live tables often lead to performance problems. It is recommended to use cached tables whenever possible.
Custom data scripts can be used to connect live to R. This is often slow.

Constant values

Not everything can be cached automatically. If you have constant parts of expressions, you can also cache manually by creating new attributes to prevent recalculation of values. This also ensures the calculations are stored in the attribute cache and no calculations are needed on opening an application.