Harness Powerful Analytics to Accurately Forecast Operational & Business Needs

From real-time processing for baseline statistics to advanced, custom real-time analytics streaming, Circonus has an extensive analytics toolkit designed for operational intelligence.



Discover Operational & Business Insights Over Time

Use Circonus to help collect and store billions of metrics. Then run rich queries against your stored data, such as percentiles, counts below X, and aggregations across streams.


Use histograms to visualize the frequency of values within a continuous stream of data. A histogram is built by allocating data values into interval bins. Collected over time, successive histograms can reveal changes within the underlying data. Assess probability distribution, find disruptions and concentrations, and discover extreme system loads or drop-offs in real-time — all with histograms.


Use heatmaps to represent histogram data in a color-coded matrix to better visualize usage patterns of a system measured over a period of time. Color points create a visual representation of the population density of a given set of samples.

Capacity Planning

Accurate Capacity Planning with Forecasts & Usage Requirements for Resources

Plan for requirements for disk space or network load, based on actual history, using linear, exponential or periodic regression tools with a click of the mouse. Or better yet, let the system automatically select the tool best suited for your data. Easily forecast capacity loads for daily or weekly periods.


Use capacity planning with a linear regression to forecast growth. The graph shows memory usage on a system. There is a memory leak, so the memory will fill up quickly. On July,7th one node will use 60G of memory.


Visualize exponential regression and use it to scale to the future. This graph shows the number of accounts is exponentially growing, so given that the current trend continues, a number of users in a given future date can be accurately predicted.


Apply account periodicity to capacity planning. Forecast trends using a seasonal predictive model with Holt-Winters exponential smoothing over weekly periods. The graph shows web request rates and makes a forecast for the next week, forecasting trend and periodicity. We can see a slight negative trend.

Anomaly Detection

Easily View Anomalies in Data when Metrics Do Not Behave as Expected

Analyze the behavior of any kind of metric and automatically create a continuously updated forecasting model of what that data should look like over time. Easily identify problems and make faster decisions to rectify errors or make system wide adjustments.

Dynamic Alerts

Identify periods of unusual or unexpected behavior in your data with anomaly detection overlays. Instead of alerting on a threshold, the pattern-based forecast becomes the “new normal” and alerts can be based on deviations from the norm.


Directly view anomaly scores or use graphs to highlight anomalous regions. Compute an anomaly score between 0-100% for each data point – when the score reaches 100, an anomaly is detected. This graph shows API request rates that are normally constant at ~100, but experience significant deviations at certain regions. Those regions are identified and marked by the Anomaly Detection method.


Recognize patterns and detect deviations from them. Monitor system performance over time to gain insight into predictable patterns as well as fluctuations from them. This graph demonstrates how the anomaly detection method adapts to period behavior and detects deviations from usual periodic patterns.

Service Level Monitoring

Efficiently Diagnose & Resolve Problems in Scalable Distributed Systems

Visualize all collected data to visualize more accurate service level monitoring graphs. View service level violations across multiple SLAs. Use actual historical data to set SLAs that make the most sense for operations and the business.


Calculate percentile values for Service Level Monitoring data across selected time windows. This histogram shows API response times, measured from inside the service in 1 minute resolution. The overlay in this example shows the 99th percentile in order to check an SLA: “99% of all queries issued within an hour/day should be serviced within 200ms.”

Inverse Percentiles

Use inverse percentiles to calculate ratios of samples that are below a given threshold value. For an SLA that states “99% of all queries issued within an hour/day should be serviced within 200ms.” can be checked by verifying that the percentage of queries served within 200ms is larger than 99%. In this graph, the inverse quantile is computed over all visible data in 1 hour intervals.

Percentile Aggregation – Min-Max

Use percentile aggregation to accurately analyze large view ranges. A min-max Percentile Aggregation shows the maximal (and minimal) values that occur within the aggregated time interval. This graph shows the range of the 99th percentile over 2 days. The 99th percentiles are much larger than the displayed graph because there are not enough pixels to show all measured values (1m resolution: 1440m/day). Averages-aggregation is performed before graphing to reduce spikes significantly.

Box-Plot Aggregation

Use box plots to accurately view aggregated data. The box-plot aggregation shows the information contained in box plots as an overlay, the min, the max, and 3 percentiles: 25%, 50% (the median), and 75%. Box plots are available for histograms as well. This example shows a 3 month slice of a proxy request rate graph with box-plot percentiles added. The highlighted original graph only shows mean values over several hours. The 25% and 50% percentiles show typical variations of the the graph and the min and max values show the full range of variation.

Graph Comparison

Combine Multiple Graphs in the Same View for Easy Comparison & Correlation

Visualize performance data comparisons in the same graph for any time period you need – daily, weekly, monthly, even yearly as long as the data has been captured by Circonus.


Historical graph comparison allows you to compare the current graph to historic values. In this example, we see a web request metric, with graph comparison overlays with 1 day, 1 week, and 1 month delayed. This view allows us to answer important questions. Did we see more traffic then last week? Did the traffic pattern change significantly from last week or last month? Here we can see the spike is always precisely at 19:00.


Compare a graph with an overlay of a different graph to find correlation. This example shows the web request rate graph compared to the request rates graphs of the DB cluster. Note, that during Oct 1-10, there were a large number of web requests that did not cause DB requests. We might want to check the web-server logs to see which pages were requested instead.


Graph your data and use our analytics overlays and predictive algorithms to get real insight for troubleshooting potential issues, capacity planning, setting cost-effective SLAs, and more.

Learn More


Reliable threshold alerting helps your team keep your systems and service quality up. Alert team members on early signs of trouble such as when data deviates from expected values.

Learn More


Programmatically adjust monitors and alerts as architecture evolves. Adjust and customize the UI and extend services with third-party tools like Python, Ruby, or Java.

Learn More

Get Started Today.
Create an Account or Get In Touch.