Cloud service polling overview

Cloud service APIs poll instances/VMs and volumes regularly to monitor details about resources, such as status, volume IDs, subnet IDs, memory usage, CPU usage, and IP addresses.

The SolarWinds Platform sends API requests (also known as "calls") to cloud services to retrieve status and metric data used to monitor resources, trigger notifications, track system health, and more. Third-party products and custom code may also send API requests for various purposes.

Note these details about cloud service polling:

• AWS and Azure both provide 1 million free API requests per calendar month. If free polling limits are exceeded, cloud accounts are charged for extra requests within the remaining time frame. See AWS polling and Azure polling.
• Cloud metrics gathered by cloud service APIs and retrieved by the SolarWinds Platform vary from OS metrics, as described next.

To avoid exceeding polling limits, consider toggling Auto Monitoring off when adding a cloud account to block polling for new instances/VMs launched and then discovered for an account. You can enable monitoring for individual instances/VMs later.

Cloud metrics vs. OS metrics

Cloud services APIs capture basic metric data for instances/VMs and volumes so you can allocate resources as needed, such as partial CPU processing and disk space across multiple instances/VMs. These resources can change through direct interactions and automation. For example, when the Amazon EC2 web service reports data to the SolarWinds Platform, it calculates the percentage of assigned resources shared between instances.

Cloud metrics differ with OS metrics due to the fluid nature of cloud computing. OS metrics directly capture values from the core system, not the assigned amounts. This data does not calculate shared resources or other users attached to the instances and volumes. This data directly displays the actual usage at a polled point in time.

Both values provide insight into potential and actual issues with performance and resources. Metrics report vastly different information to the cloud and OS based on how allocated resources and metric calculations.

CPU steal is an example of cloud vs. OS metrics. When CPU usage and metrics spike in a cloud environment, multiple processes and instances/VMs in the cloud may access the CPU as multiple owners. Typically, OS metric spikes tend to look like noisy neighbors. The cloud metric data better represents the data as shared resources usage across multiple owners with metrics broken down by owner.

To better define resource usage and alerts, SAM and integrated VMAN display cloud instance metrics throughout all cloud resources in SolarWinds Platform Web Console views, resources, hover-over data, and reports. Cloud metrics, including calculated health status, CPU load, and IOPS data, are used to apply global cloud thresholds that trigger alerts and status changes.

For instances and VMs managed as nodes, the SolarWinds Platform pulls specific OS data for memory and provides additional data through Orion agent, WMI, and SNMP polling methods.