VictoriaMetrics takes organic growth over investor pressure

Interview Monitoring biz VictoriaMetrics is relatively unusual in its field. It is yet to accept external investment, preferring instead to try to grow organically rather than being forced to through a private equity meat grinder by committing to grow by X every year until the investor exits.

Co-founder Roman Khavronenko, who was speaking to us at a recent Kubecon event about open source, licensing, and the pressure of accepting venture capital, opines that, barring a few success stories, "in most cases, startups fail because of this huge pressure."

In 100 percent of license changes, it's only a matter of money

Khavronenko and his colleagues founded VictoriaMetrics in Kyiv, Ukraine, in 2018, although it is now headquartered in San Francisco. The company operates open source time series database monitoring. Customers keeping tabs on a system's health is essential in spotting minor problems before they become major. Monitoring solutions like VictoriaMetrics' product both record and query telemetry and metrics from a variety of sources.

Talk of avoiding pressure from investors inevitably leads the conversation to the recent moves towards license changes in the open source world. Examples include HashiCorp's shift to the BSL. Khavronenko's tells us: "In 100 percent of license changes, it's only a matter of money."

VictoriaMetrics uses an Apache 2 license for its open source product, but the enterprise tooling is resolutely closed source and requires an activation key.

The open source product, he says, is full featured. The company claims it can store millions of data points per second on a single instance or scale across multiple datacenters. It's simple to set up – just a single binary and K8s operator will do the trick – and is compatible with Prometheus and Grafana.

Khavronenko says: "We are fully committed to using Apache 2, and we are not going to change that. And most of our work – probably 99 percent of what we do every day – is about open source."

VictoriaMetrics' enterprise incarnation adds some extra functions, including improved alerts, machine learning for anomaly detection, and Kafka integration. Perhaps most importantly for enterprise users, it also increases security and support directly from the company rather than the community.

The enterprise product is currently being offered under a 60-day free trial.

The company says it remains enthusiastic about open source, and values the community that has followed it from its beginnings on GitHub to where it is today, Khavronenko wants us to know.

He says VictoriaMetrics is "community-driven because community is something which gives you real insight about how this tool can be used in the real world to solve the real problems."

Khavronenko also praises the feedback it gets from the community but admits that while VictoriaMetrics is receiving plenty of feature requests, it can't accept them all. Not even from big companies with potentially deep pockets. And particularly not if that feature has a minimal use case unlikely to benefit the rest of the community.

For the future, the co-founder sees potential in AI and machine learning – the company already has some machine learning in its anomaly detection, which was trained on historical time series data.

• IBM's vintage Db2 database jumps on AWS's cloud bandwagon
• Microsoft, Databricks double act tries to sew up the data platform market
• Microsoft touts mirroring over moving in data warehouse gambit
• Developers, developers, developers: How 'serverless' crowd dropped ops like it's hot

Khavronenko is skeptical about the abilities of certain generative AI tools in being able to create, for example, SQL queries from natural language – "if your language is too complex to use," he says, "probably you need to pick a better language..." However, Khavronenko also notes the technology has great potential when it comes to monitoring.

"Where AI can help in monitoring is to do things that a human can't do," he adds.

A human can handle a few signals, but when dealing with the hundreds of millions or even billions of signals emitted by services such as Kubernetes clusters, an AI can be used to spot patterns that a human might miss. Or perhaps tell the human that they're collecting too much data.

However, as Khavronenko notes, the amount of data collected continues to grow. "So far," he says, "I haven't seen any case where the amount of telemetry data decreased. It is always increasing..." ®