SolarWinds Platform server performance

Beginning with the NPM 10.2 release (Fall 2011) and including all products built on the SolarWinds Platform, SolarWinds optimizes monitoring performance by automatically balancing polling rates and polling intervals in heavy polling environments.

To accomplish this balance, when required, SolarWinds automatically extends configured polling intervals to decrease the overall polling rate in your environment. To determine when this polling rate adjustment should occur and when users should be notified, SolarWinds has established two polling rate thresholds: a polling warning level and a maximum polling limit level. These levels are based on an internal calculation that includes both the number of network objects polled and the configured polling interval.

As a result of these polling enhancements, it is no longer necessary for a user to compensate for performance issues by manually setting polling rates higher than required to ensure that data is actually polled as frequently as desired. All polling jobs will be completed. If the polling rate is more than the server can handle, the polling intervals are automatically increased to handle the higher load.

Poller Performance Measurements

Much of the SolarWinds Platform server performance measurement and management can be done directly from the Orion web interface. In order to understand how well a poller is performing, you must first understand where to look for poller performance data and how to interpret that data. In the Administrative Settings of the SolarWinds Platform Web Console Settings > Polling Engines you see a variety of polling measurements. The poller specific fields represent aggregated measurements of all active pollers. Below is a brief description of the polling engine data related to Orion performance.

Polling Completion

Polling completion represents the percentage of polls across all pollers that were completed without the need to delay polls. The new Orion poller can delay polls if it detects that the SolarWinds Platform server resources are overtaxed. To make it easier to detect ongoing polling issues, the polling completion number is an average of the last one hundred polling jobs. A polling completion number less than one hundred only indicates delayed polls, not failed polls. Polling rate should normally be about one hundred. Significant declines in polling rate are typically due to the SolarWinds Platform server memory or CPU constraints.

Polling Rate

The polling rate indicates the percentage of maximum polling rate possible for your server. This number is derived from the Total Job Weight value and the polling interval settings. Each polling job is assigned a weight to best estimate relative load the job places on the server. If the polling rate reaches eighty five percent or greater the SolarWinds Platform server will give a poller status warning popup. At this point you should add an additional poller or scale back you polling rates if you have increased the rates from the default settings.