Jun 13Vibe with Hermes Agent — Bengaluru · RSVP
ToolsMCPBlogResearchCommunityStar on GitHub
All Tools
O
DataFreeOpen Source

OPENTSDB

Distributed time-series database on Hadoop/HBase

LGPL-2.1

ABOUT

AI infrastructure generates billions of time-series data points — GPU utilization across thousands of nodes, model serving latency percentiles, embedding cache performance, training job resource consumption — that need to be stored, queried, and analyzed historically. OpenTSDB provides a purpose-built time-series database on top of HBase that scales horizontally to handle millions of metrics per second with automatic downsampling, aggregation, and retention policies. Its HTTP API and telnet protocol are compatible with StatsD collectors and the wide ecosystem of metrics tooling, making it a proven backend for large-scale ML observability.

INTEGRATION GUIDE

1. Store and query GPU utilization metrics across a thousand-node training cluster with sub-second query latency 2. Track model inference latency percentiles (p50, p95, p99) over months of production data 3. Aggregate metrics from StatsD collectors across distributed model serving infrastructure 4. Downsample high-cardinality metrics automatically for long-term trend analysis and capacity planning 5. Detect performance regressions by comparing current metric patterns against historical baselines

TAGS

monitoringtime-seriesdatabasehadoophbasemetrics