Apache Doris unifies logs, metrics, traces and AI agent events on a single high-performance analytical foundation, so teams troubleshoot faster, control costs and keep improving AI quality over time.
When teams unify observability across logs, traces, metrics and AI agent events, five things shift at once: incident detection, user experience, cost at scale, AI quality, and the link between system behavior and business outcomes.
Observability connects logs, traces, and metrics to help teams detect anomalies and find root causes before failures spread. For AI agents, it turns LLM timeouts, prompt failures, failed tool calls, and runaway loops into analyzable execution data.
Three teams run Apache Doris as the analytical foundation for observability: at scale, on real-time operational data, across logs, metrics and events.
“After migrating from Grafana Loki, Apache Doris now powers the logging system across all of MiniMax's business lines, serving PB-scale data with over 99.9% availability and second-level query latency on hundreds of millions of log entries.”
“NetEase migrated its Eagle monitoring platform off Elasticsearch and its IM time series platform off InfluxDB, consolidating both workloads on Apache Doris for faster queries, lower storage costs, and a more flexible index design.”
“The shift from Elasticsearch to Apache Doris has slashed storage costs by 80% while boosting write performance by 4×, with inverted indexes powering full-text search and aggregations in a single SQL query.”
Five things a modern observability platform has to be good at, and the specific Apache Doris capabilities that meet each one.
Modern observability data streams from applications, Kubernetes, APIs, gateways, databases, and AI systems. Your platform needs to absorb high-volume logs, traces, metrics, and AI agent events with low latency, so every signal — from LLM requests and tool calls to RAG retrievals, token costs, evaluations, and user feedback — is ready to query in near real time.
Logs, traces, tool outputs, model responses, and agent events often arrive as nested JSON with constantly changing fields. The platform must handle schema evolution and make any new field immediately queryable, so teams can filter, investigate, and analyze new signals without rebuilding pipelines.
Observability data contains massive volumes of searchable text, from error messages, stack traces and log lines to prompts, model responses, tool outputs and agent failure reasons. Teams need fast keyword and full-text search so they can find failures, trace requests and investigate agent behavior at scale.
Observability is not only search. Teams need to slice, aggregate, and drill into massive telemetry datasets to analyze reliability, cost, and AI quality — from P99 latency and SLA trends to token usage, model cost, RAG quality, and agent task completion. That requires a high-performance analytical engine alongside fast search.
AI agent observability requires more than keyword search over logs. Teams need to correlate structured metadata with full-text and semantic search across prompts, responses, tool calls and traces, so they can find exact matches, similar failures, hallucinated outputs and recurring agent behavior patterns in one query.
Apache Doris ingests high-volume telemetry from Kafka, CDC pipelines, and streaming APIs with low latency. Logs, metrics, traces, and AI agent events become queryable in near real time for fast debugging, monitoring, and cost analysis.
