Nutanix leverages a method for determining the bands called 'Generalized Extreme Studentized Deviate Test'. A simple way to think about this is similar to a confidence interval where the values are between the lower and upper limits established by the algorithm.
Another web source3 shows an example of how anomaly detection works in Nutanix Prism Central. In the video, you can see that when the observed value of a metric deviates significantly from the predicted value based on historical data, an anomaly event is triggered and displayed on a chart.
Therefore, by comparing the observed values with the predicted values based on historical data, Nutanix anomaly detection can identify abnormal behavior and alert you accordingly.

According to the image description, the exhibit shows a screenshot of a computer screen with a table of data. The table has 6 columns and 4 rows. The columns are labeled "Date", "Time", "Status", "Details", "Configuration", and "Duration". The rows contain information about different configurations and their statuses. The background is a light blue color. There is a pop-up message on the bottom right corner of the screen that reads "This question has been answered. Click here to enable the edit button." Based on the table data, it seems that the screenshot is taken from the Prism Central web console, showing the results of a cluster health check report. The report shows that there are four failed checks: Host Password Policy, CVM Password Policy, NTP Configuration, and Cluster Time Skew2. These checks indicate that some initial cluster configuration tasks were missed during the deployment process.
The Host Password Policy check verifies that the host passwords are not set to default values and are compliant with the password policy defined in Prism Central3. The CVM Password Policy check verifies that the CVM passwords are not set to default values and are compliant with the password policy defined in Prism Central4. These checks are important for ensuring the security and compliance of the cluster. Therefore, changing the host and CVM passwords from their default values and following the password policy guidelines are two initial cluster configuration tasks that were missed.

A GPU (graphics processing unit) is a specialized hardware device that can accelerate graphics rendering and computation for applications that use APIs such as DirectX, OpenGL, CUDA, and OpenCL. A GPU can also offload the CPU from encoding and decoding tasks for remote display protocols such as Frame Remote Desktop Protocol (FRP). A VM can use a GPU either by directly accessing a physical GPU (pGPU) on the host or by using a virtual GPU (vGPU) that shares a pGPU with other VMs. A user running a computer aided design (CAD) application may benefit from using a GPU or a vGPU to improve the performance and responsiveness of the application, especially when moving windows or rendering images. However, if the GPU or vGPU is not properly configured or provisioned, the user may experience slow response time within the VM. Therefore, to diagnose the problem, the administrator should monitor the GPU Usage metric for the VM. The GPU Usage metric shows the percentage of GPU resources that are consumed by the VM over time3. The administrator can use Prism Central to view the GPU Usage metric for each VM in a chart or a widget4. The administrator can also use Prism Central to view other metrics related to GPU performance, such as GPU Memory Usage, GPU Encoder Usage, and GPU Decoder Usage3. By analyzing these metrics, the administrator can determine if the VM is using the GPU efficiently and optimally, or if it needs more or less GPU resources.

Explanation
https://portal.nutanix.com/page/documents/details/?targetId=Web-Console-Guide-Prism-v50:wc-block-services- Provision storage on the Nutanix cluster by creating a volume group. Create a client whitelist to enable access to the volume group by using the IP addresses or client initiator IQNs in a whitelist (as part of the volume group configuration). Create a secret for the volume group if you are using CHAP authentication.