Last November, a good friend of ours, Paul Dix from InfluxData, invited me along to their InfluxDays summit in San Francisco. (Sidenote: Our very own David Cromberge will be presenting at InfluxDays summit in New York next month on why NOT to build your own time-series database and our evolving architecture behind the scenes)
At the event, I met another veteran in the monitoring space, Baron Schwartz from VividCortex, who I've gotten to know at a couple of events. Baron is super smart and has enough experience with monitoring to know when to call BS on certain fads in our industry, the main one being Anomaly Detection.
We've also seen Anomaly Detection being requested as a feature in deals, and we've seen competitors add it, and then later heard it's terrible, mostly for the reasons Baron presents in his talk. I still remember taking Skyline and applying it to one of our customer's metrics, and turning 100,000 metrics into 10,000 anomalies. It just created more noise from the noise.
So I wanted to post this video here for all the users that want us to do Anomaly Detection because alternative solution XYX does it. Yes they do it, but it doesn't work, and we believe there's better approaches to cutting down the noise than trying to build something that doesn't seem to work in practice.
Here's Baron's talk from InfluxDays: