With the release of ArcGIS Velocity in ArcGIS Enterprise 12.1 on Windows and Linux, the ArcGIS Velocity product team conducted a set of system tests to evaluate existing throughput capacity to evaluate performance.
The goal of this test study is to help organizations understand how ArcGIS Velocity performs under defined conditions in ArcGIS Enterprise and on specific AWS infrastructure. Over time, the study will be expanded to include additional cloud providers, hardware configurations, and test scenarios that further exercise feeds, real-time analytics, and outputs such as feature layers. This continued effort will provide broader context for how ArcGIS Velocity ingests, analyzes, and distributes real-time streaming data in ArcGIS Enterprise.
For this initial release, and until otherwise noted, the results of this study are intended to illustrate the expected performance of selected feeds and real-time analytics across different hardware and operating systems.
While you are encouraged to compare these findings with your own planned deployment of ArcGIS Velocity, the results and metrics presented here should be interpreted within the specific context of the testing environment and not generalized without consideration. Variations in hardware, data volume, data size, analytic complexity, and other factors can significantly influence performance in your own deployment.
The objectives of these performance tests was to determine the following:
Determine the maximum sustained throughput (features per second or f/s) for Kafka, HTTP Receiver, and HTTP Poller feeds
Determine the maximum number of a Kafka feeds running at a throughput of 1,000 f/s can be run simultaneously
Determine the maximum sustained throughput (f/s) to a stream layer in a real-time analytic
Determine the maximum sustained throughput (f/s) through a reference real-time analytic consisting of different tools to a stream layer
To make best use of this test study, it is recommended to review the following ArcGIS Velocity resources: