Overview

SixthSense Observability Premium users have the following benefits including the ones available for the SixthSense Observability Standard application. For more information about the standard benefits, see Overview.

The following features are available only for SixthSense Observability Premium users.

Forecasting

This feature provides insights into future trends of application metrics using an advanced algorithm powered by Transformers. This algorithm combines multi-horizon forecasting and interpretable analysis of temporal dynamics to deliver accurate predictions. The machine learning algorithm treats each received metric value as a data point and learns from a historical dataset, considering both static and dynamic events from the past. It constructs forecasts by interpreting future data points. To enhance accuracy, the model takes into account a short history of time-series data. This enables it to detect metric patterns associated with specific events, even if they occurred in previous years. The algorithm forecasts similar seasonality in the current year around the same time. The ML models are trained on historical data, which can be provided by the user in bulk, and they continue to learn and improve through model retraining. Retraining can occur either as a batch process or at shorter intervals based on license purchased.

The current machine learning (ML) models are trained using application performance metrics listed in the Benefits section. These models have the capability to forecast up to a day.

Forecasting value

Forecasting with ML models in infrastructure management is beneficial as it enables continuous learning, accurate resource utilization forecasting, cost reduction, performance optimization, and proactive issue detection. Unlike manual calculations based on past data and significant events, ML-powered forecasting provides more precise predictions in various scenarios, resulting in improved efficiency and cost savings.

Dynamic baseline

Dynamic baselining is a forecasting approach that emphasises the confidence interval, incorporating data variability and uncertainity. By replacing static benchmarks/thresholds with adaptable thresholds, it delivers accurate performance expectations. This method minimises false positives, enhances anomaly detection, and enables proactive management. With a focus on the range of expected values, dynamic baselining empowers data-driven decision making for optimal system performance. It provides a flexible framework for monitoring and responding to changing conditions, ensuring efficient resource allocation and maintaining service excellence.

Dynamic baseline value

This feature replaces the manual setting of alert thresholds for thousands of metrics. By automating this process, it saves time and effort while ensuring consistency and accuracy. Additionally, the feature serves as an adaptable benchmark that considers seasonality in anomaly detection. Overall, the feature enhances operational efficiency, reduces manual workload, and provides more reliable performance monitoring in dynamic environments.

Anomaly detection

The Anomaly detection feature detects breaches of the real-time metric. The algorithm used for anomaly detection is an algorithm powered by Transformers. Whenever a metric breaches the dynamic baseline, the severity of the breach is calculated by the algorithm and is marked distinctively in three colours: Red (Critical), Orange (Major), Yellow (Minor). The metrics ingested by the application in this release are listed in the Benefits section. These golden metrics are related to an application’s performance.

Anomaly detection value

Detecting anomalies in essential performance metrics and its various types are not possible without this feature. The current release solves the problem of monitoring various anomalies in any number of metrics related to an application performance. The severity of the anomaly defines the criticality and can avoid false positives.

The benefits for SixthSense Observability Premium customers are as follows:

Benefits

Following capabilities with their application performance metrics are available for premium users. This feature helps in forecasting anomalies based on the previous values.

APM

• Load
• Error
• Response time

VM

• CPU usage
• RAM available
• Storage available
• Network bandwidth usage

Kubernetes monitoring

• CPU usage
• Memory usage
• Total fs usage