I want to share this with you all, Luke Wignall and his team from NVIDIA have created some great AppGuides, that helps with understanding how many users can you put on a NVIDIA GRID system with a K2 in a VMware environment. These guides are made together with vendors such as ESRI. The guides focus on following apps ESRI ArcGIS Pro.
In this blogpost my goal is to highlight the great work NVIDIA have done creating the scalability app guides and these guides helps you if you want to virtualize ESRI Pro with NVIDIA GRID and VMware Horizon. The guides are great – cause they give an idea what you would require in a virtualized environment and these can be reused for other virtualized platforms such as Citrix and Microsoft – keep in mind that results might be different. If you would like to get more informations about how the setup is configured and which methodology i recommend you read the AppGuide, you can download it in under source in the end of this article.
The appguides gives a great idea to understand the impact of CPU and how the GPU are giving value.
About ESRI ArcGIS Pro
ESRI ArcGIS Pro 1.0 is a Geographic Information Systems (GIS) application for mapping, visualizing, editing, and analyzing spatial data. Esri recommends a GPU for best end user experience, but as ArcGIS Pro 1.0 also generates heavy CPU load, this also needs to be considered in architecting your vGPU solution. The size of your map data, the concurrency of your users, and the level of interaction with 3D data all need to be considered when defining your user groups.
Results NVIDIA Appguide for ESRI ArcGis Pro
The following are the full results of our testing. The baseline was the 45 second draw time sum – anything greater than that value represented a worsening UX while less would be a better UX. Looking for both performance and scalability, we tested looking for the greatest number of virtual desktops, and therefore the greatest scalability, while still within performance expectations and the threshold of 45 seconds. It’s important to note that your users, your data, and your hardware will impact these results and you may decide a different level of performance or scalability is required to meet your individual business needs. Tests were also run to look for potential NUMA issues that can negatively impact performance. This is where the physical GPU and its PCI-e channels are tied to one physical CPU, while the virtual desktop is running on the other physical CPU, so communication with the physical GPU has to move over the QPI between the two physical CPUs. This creates a bottleneck and can cause performance issues. However, in our testing, the application is sufficiently CPU bound that NUMA affinity made little difference. The results in the table below show the decrease in performance as we increased vCPU counts, and then the increase in scalability with synthetic human behavior (think time):
ArcGIS – users per server
Based on NVIDIA Performance Engineering Lab findings, NVIDIA GRID provides the following performance and scalability metrics for Esri ArcGIS 3D Pro 1.0. These metrics are based on tests with the lab equipment shown in the graphic below, using the Esri API based “heavy 3D” benchmark and in working with Esri to determine acceptable performance. Of course, your usage will depend on your models, but this is guidance to help guide your implementation.