ArcGIS Dashboards is known for its great map user interface and ease of design and deployment. After successfully building the ADOT bridge dashboard with 13,000 point features in ArcGIS Dashboards, we went on to build a dashboard for pavement segments with over a quarter of a million line segments, and the performance was unbearable – switching from the initial statewide extent to the extent of an MPO took over 20 seconds. Yuck!
Determined, the ADOT DataViz team decided to use all tools at our disposal to optimize the dashboard in several areas discussed below.
Layer Preparation
We first tackled data preparation. We reprojected the data to Web Mercator, removed unnecessary columns, and generalized the data to 1m resolution. We noticed only a slight improvement in response times after implementing these changes.
Layer Registration
Next, we tackled server settings on each published feature dataset. When we applied read-only privileges and increased the cacheControlMaxAge from zero to two weeks we measured a 20% reduction in response time! We were finally getting somewhere!
Server Configuration
We then experimented with user-managed and ArcGIS-managed databases. We built our first dashboards with data stored in our SDE database that we published to Portal as user-managed feature layers. We achieved a huge performance improvement after publishing the data as ArcGIS-managed hosted feature layers. Combined with all previous steps, publishing the data as ArcGIS-managed hosted feature layers reduced the Pavement dashboard response time from 20 seconds to under 3 seconds!
Architecture
In the last leg of our optimizing journey, we explored 3 different architectures: Portal, ArcGIS Online (AGOL), and a Portal/AGOL hybrid.
With the Portal architecture, we built the dashboards using data stored in an SDE database on our server. With the AGOL architecture, we built the dashboards using feature services published to AGOL. The AGOL architecture resulted in slightly faster performance than the Portal architecture and superior map rendering functionality. However, data storage for the entire family of dashboards would consume approximately 70 AGOL credits per month! We found the hybrid architecture to be a happy medium – Portal-hosted data consumes no AGOL credits. Adding the REST services from Portal to our AGOL account allows us to use the superior ArcGIS Dashboards Beta, which is currently available only in AGOL.
In short, we found that properly preparing datasets by using the Web Mercator projection and removing unnecessary columns, then publishing the data with read-only privileges with 2-week cacheControlMaxAge as ArcGIS-Managed hosted feature layers drastically improved dashboard performance, and we were able to avoid consuming AGOL credits by hosting the data in Portal and using the REST services to build the dashboards in AGOL.