Solutions

Sentinel Smart Alerting

Back Solutions

When your key technology asset is a highly-available, easily-scalable system build on a distributed architecture you must not overlook a risk of something going wrong causing performance issues. This is why monitoring a system in production is mandatory.

But what if you would have a smart monitoring system keep the watch for you? Better yet, what if the system would have an AI engine that would enable a “sixth sense”, a premonition of something going wrong in the future?

Sentinel does all that and more. Sentinel detects when anomalies are about to happen and creates a SMART alert (email or SMS alert enriched with system metric diagnostics snapshot and a cause of error, and a hint from algorithm that constantly learns what caused this problem). Our monitoring solution saves relevant parameters such as memory, CPU, network, disk statistics, and application metrics of interest. Some of the common possible causes of production performance drop can be disk latency, network issues or even a background process in the database. This is where our machine learning algorithm comes into play. By analyzing the multiple parameters in real time it stays on the lookout for anomalies that could lead to greater issues down the road.

Sentinel Smart Alerting system is build using Apache Kafka, Apache Spark (Streaming and MLib), Apache Cassandra, InfluxDB, Riemann and Telegraf. The following graph explains how Sentinel fits with an existing system.