Sematext releases anomaly detector for Apache Spark

Sematext is a company working deep within the core of the web.

As we’ve reported before, it has products that collect the logs from across servers; it has other products that intelligently detect anomalies in a site’s performance. The company just released a new version of its SPM product, making it the first anomaly and performance monitoring system for Apache Spark.

“By releasing the first Spark monitoring product to market with SPM, we have just filled a big hole in the Spark ecosystem,” Otis Gospodnetic, the company’s founder and CEO said.

In recent years, Spark, an open-source  software platform for data analytics, has grown in popularity within the Apache ecosystem. Apache is one of the main software systems used for running servers. It’s used by large companies like Yahoo and Intel, according to a 2013 article in Wired.

Gospodnetic explained, when we previously visited, that the core of SPM’s value is its ability to intelligently detect anomalies. That is, it establishes a baseline for metrics and redefines anomalies as those baselines change.

The easiest example to give is a spike in traffic. If a site normally gets 30,000 views per month, a 5,000-view day would be an anomaly; it would not, however, be an anomaly if the site gradually grew to 150,000 views per month. Rather than requiring development teams to define anomalies, SPM defines and redefines them based on activity. Sematext’s latest SPM release can now apply the same tools to any instances of Spark that a client is running across its servers.