OpenTSDB

OpenTSDB is a distributed, scalable time-series database built on Apache HBase for storing and querying massive volumes of metrics without losing granularity.

Reviewed by 7wData

On this page

Publisher review

OpenTSDB is a distributed, scalable time-series database built on Apache HBase for storing and querying massive volumes of metrics without losing granularity. Created in 2010 at StumbleUpon to monitor billions of data points per day, it evolved into a production TSDB used by high-scale infrastructure operators including Yahoo (22M+ writes/sec), Box, Cloudflare, and Salesforce. The system runs as a cluster of Time Series Daemon instances backed by HBase and ZooKeeper, handling long-term metric retention indefinitely and supporting queries across years of historical data.

OpenTSDB remains in active use at companies managing enormous metric volumes, particularly in financial services, social networks, and cloud infrastructure teams. However, it has fallen out of favor for new deployments due to the operational burden of its Hadoop/HBase/ZooKeeper stack, performance degradation under sustained high load (writes drop from 2.5M to 500k points/sec under stress), and the need for pre-aggregation to query datasets spanning days or weeks. The project entered maintenance mode in 2021 and has not shipped a major release in over three years, though GitHub shows ongoing community contributions. Modern alternatives like Prometheus, InfluxDB, and VictoriaMetrics offer simpler deployment models and comparable performance for most new use cases.

Get the AI & data signal, daily.

335k+ subscribers read this every morning. One email, both newsletters. Unsubscribe anytime.

How it works

  1. Distributed time-series storage on HBase

    Stores trillions of data points across a Hadoop/HBase cluster with automatic sharding and replication for fault tolerance.

  2. HTTP-based query API

    Read and write metrics via HTTP GET/POST requests without requiring a custom driver or protocol implementation.

  3. Custom metric tagging

    Organize metrics with arbitrary key-value tags (metadata) for flexible filtering, aggregation, and dashboard organization.

  4. Aggregation and mathematical operations

    Perform sum, avg, min, max, count, and other operations on metrics within the query layer, reducing data transfer.

  5. Multi-datacenter replication

    Replicate metrics across data centers via HBase replication for disaster recovery and geographic distribution.

  6. Horizontal scaling without resharding

    Add new TSD instances to the cluster to increase write throughput and query concurrency without migrating existing data.

  7. OpenTSDB Horizon visualization UI

    Built-in web dashboard for graphing metrics, generating alerts, and exploring historical time-series data.

Strengths and trade-offs

Strengths

  • Proven at massive scale: Yahoo, Box, Cloudflare, and others run petabytes of metrics through OpenTSDB in production.
  • No data loss on metric granularity: retains raw data indefinitely without forced downsampling, enabling long-term forensic analysis.
  • True horizontal scaling: add TSD instances to cluster without resharding or migrations; queries fan out automatically.

Trade-offs

  • Operational complexity: requires running and tuning Hadoop, HBase, ZooKeeper, and OpenTSDB in concert; teams report it became 'a headache' after 6 months in production.
  • Performance degradation under sustained load: write throughput drops from 2.5M to 500k points/sec under sustained stress; query timeouts common on datasets spanning days without pre-aggregation.
  • Maintenance mode with no major releases since 2021: project fell out of favor relative to Prometheus and InfluxDB; community support narrower for operational troubleshooting and new features.

Pricing context

OpenTSDB is free and open source under dual licensing (LGPLv2.1+ and GPLv3+). There is no commercial version, managed cloud offering, or subscription. Costs are limited to self-hosted infrastructure: Hadoop/HBase cluster maintenance, ZooKeeper coordination, and operational staff to tune performance and manage region compactions.

Alternatives

User reviews

No user reviews yet. Be the first to write one.

Sources

Reporting on this tool draws on these publicly available sources.

  1. github.com — Project status, current version (2.4.1), GitHub activity, 5.1k stars, 1.2k forks, open-source license (LGPLv2.1+/GPLv3+)
  2. medium.com — Real-world performance trade-off: query timeouts on 7+ days of data, need for 60x speedup via rollups, noisy neighbor problem in HBase RPC queues
  3. medium.com — Operational complexity: HBase compaction failures with millisecond timestamps, region split cascades causing heap exhaustion and GC pauses
  4. github.com — Production users: Yahoo (22M writes/sec), Box (10M), Turn (4.5M), Tumblr (1M QPS), Pinterest (60B+ data points), eBay, Cloudflare, Salesforce, Hulu
  5. opentsdb.net — Official homepage, founding at StumbleUpon in 2010, feature set, API documentation
  6. opentsdb.net — Founding date (2010), original creator Benoit Sigoure, maintenance mode status since 2021
  7. www.alibabacloud.com — Comparison of OpenTSDB limitations vs. Prometheus and InfluxDB, operational complexity assessment
  8. db-engines.com — DB Engines ranking and historical trend data showing OpenTSDB declining in popularity relative to modern alternatives