The pace at which technology is evolving these days, it’s challenging to monitor the modern infrastructure. The volume, variety and velocity of data, the change in business dynamism and associated scalability, growth, M&A makes things complex.
Anomaly detection reduces the complexities associated with monitoring and managing performance of the infrastructure and applications that drive revenue and profitability. With anomaly detection we can focus on identification of data points, situations that are not ideal to the desired pattern.

Using Anomaly detection momentary problems can be detected along with actionable insights into specific incidents within the infrastructure (server, application, database, other components).
The value of anomaly detection significantly increases when end-to-end monitoring is enabled (client-side, infrastructure side, as well as logs). Anomaly detection also provides actionable insights into business anomalies, such as, Order, revenue, etc. Machine learning, Artificial Intelligence, and analytics are integral components of anomaly detection.

Types of anomalies

Anomalies are broadly classified into three types

Point Anomalies

A dataset is said to be Point anomaly if it is too far off from the rest of the data.

Contextual Anomalies

When a group of data points significantly deviates from the rest of the data points but in same context, it is considered as an anomaly.

Collective Anomalies

Combination of many instances together considered as an anomaly.

What It does and why

Anomaly detector keep an eye on streaming data and compares with the baseline. Anomalies are generated if any pattern breaches the baseline in the production environment. Such breaches may be notified in real-time with detailed insights into the root cause of an anomaly.
Predictive Analytics in anomaly detection is highly productive in terms of identification of faults, strange pattern in large and complex datasets. It’s been found crucial for the systems which require smooth and secure operations.

How Cavisson Alerting mechanism helps in identifying anomalies

Cavisson software monitors each layer of application stack and collects the real-time streaming data. All monitoring data is stored in a proprietary big data engine, that allows complex computing algorithm for detailed analysis in real-time and combines it with advanced machine learning to detect anomalies at the earliest.
Cavisson pioneered ML-driven Load index-based alerts to reduce false alarms.

Detect anomalies using

Dynamic baseline
Analyzes real-time streaming data to determine performance baselines automatically and dynamically using system intelligent algorithms which learns how the application performs in different scenarios like under load and in normal performance. Its starts immediately without any thresholds / rule configuration.
Load index baseline alerts
An advanced type of baseline alert which works upon the load on the system instead of baseline trend. The baseline is predicted by learning the behavior of the system with load and then utilizes the learning to compare the current data at current load with the baseline value at current load. Monitoring the application behavior, it’s been concluded that the response time of the application is proportional to the load on the system. Which means the increase in the load results to increase in the response time. Therefore, at a point of time the response time reaches beyond the threshold value, ideally there must be an issue and should generate an alert.
Predictive alerts
Technology that allows us to predict certain events in the production environment. It is based upon machine learning algorithm which learns from the collected data over a period of time and makes predictions based on the data. Predictive alerts not only collect and report on the incoming data but also search for relation between the individual data points to accurately predict the future trends.
Correlated alerts
The method of grouping highly related alerts into one incident to increase the importance of occurred incident. Correlation of alerts is done on the basis of context, topology, time. It collects all the data from the integrated monitors and correlates it to detect the root cause of the incident.
Cavisson software automatically identifies hotspots where application is taking more time in execution and hence slowing down the response to the user. It automatically identifies slow transactions and takes thread dumps so that DevOps team can identify where application was stuck or was slow in execution.


Monitor performance across enterprise layers
Rapid identification of potential performance threats
Detection of unusual behaviors across data points
Find anomalies in event stream using historical as well as real-time data
Real-time analytics and diagnostics
Eliminates manual effort to provide efficient results
Get alerted before performance affects revenue or the operational efficiencies