- Blog /
- How DreamHost Slashed Memory Usage by 80% and Scaled to 76 Million Time Series

For any growing business, there comes a point where the tools that once worked perfectly begin to show their limits. This is especially true for monitoring infrastructure. As your user base, services, and data volumes expand, the pressure on your monitoring stack intensifies. For web hosting leader DreamHost, with over 1.5 million websites to manage, their existing open-source solutions simply couldn’t keep up.

They found their monitoring stack would “fall apart” under the load of high-cardinality data, consuming immense resources just to stay operational. The challenge was clear: find a solution that could handle massive scale without the operational overhead.
After evaluating the usual suspects, DreamHost chose VictoriaMetrics, and the results speak for themselves. The switch wasn’t just an incremental improvement; it was a transformative one.
Here are some of the key gains they experienced:
The ultimate goal of monitoring isn’t just to collect data; it’s to gain visibility that drives business value. For DreamHost, the stability and efficiency of VictoriaMetrics led to a “massive improvement in visibility” into what is happening with their customers’ websites. This allows them to be more proactive, improve service quality, and build a better customer experience.
DreamHost’s success is a powerful example of what happens when a great product meets a challenging problem. If their story resonates with you, here’s how you can learn more:
VictoriaMetrics Anomaly Detection has had a productive year with lots of user feedback that has had a major impact on product development. We’ve added improvements across the board: in core functionality, simplicity, performance, visualisation and AI integration. In addition to bug fixes and speedups, below is a list of what was accomplished in 2025.
January 2026 updates deliver quality of life improvements, performance optimizations, and tighter Kubernetes integration across the VictoriaMetrics Observability Stack.
Cluster mode in VictoriaLogs is not a separate build. It is the same victoria-logs binary started with different flags, so you can scale out without a migration step. Storage nodes persist data on disk, while gateway nodes can stay stateless by pointing to storage with -storageNode. It also ships with practical safety switches, like read-only protection when -storageDataPath runs low and optional partial results when a storage node is down.
In the last quarter of 2025, VictoriaMetrics Cloud brings many great features: New powerful Capacity Tiers, the expansion to the us-east-1 (N.Virginia) AWS region in the US, new Notification Groups, a Terraform provider to complete your IaC, 9 brand new Alerting Rule Integrations and much more.