- Products
- Solutions
Real-Time Analytics
Real-time database for real-time data warehousing, customer-facing and agent-facing analytics
Lakehouse Analytics
The fastest lakehouse SQL engine, replacing Trino/Presto, SparkSQL
Observability in the AI Era
The most cost-effective alternative to Elasticsearch observability
VeloDB for AI
The AI-Ready analytics database for the AI era
- Docs
- Resources
- Pricing
- Contact us
Use Case
Observability in the AI Era
Beyond Infrastructure
Traditionally, observability platforms have focused on the "Three Pillars"—Logs, Metrics, and Traces—to decipher the internal states of deterministic systems. However, the rise of LLMs and AI Agents introduces a new paradigm of non-deterministic outputs and the challenge of hallucinations. In this new era, observability must transcend the physical layer. It is no longer just about maintaining system stability, it is more important to improve the accuracy,reliability, and cognitive effectiveness of AI applications.
Cost Efficiency
High costs for ingesting and storing vast data volumes, while keeping query performance, are a barrier set to rise with business growth.
Semi-Structured Data
JSON in logs and traces needs flexible schemas that also support high-performance storage and analysis.
AI Effectiveness
As LLMs and AI agents introduce non-determinism, observability must evolve to safeguard accuracy and fuel the continuous improvement cycle.
Why Choose VeloDB
10x Cost Effective Compared to Elasticsearch
Cut storage volume by 80% while keeping inverted index.
5x ingestion rate with inverted index
2x full-text search performance
Efficient and Flexible JSON Data
VARIANT data type extract JSON fields as sub-columns without ETL
10:1 compression ratio, 3x compared to text or binary JSON
8x analytical performance
Ready for AI Observability
AI Observability for agents and applications with Langfuse-doris
Infra Observability with Grafana, OpenTelemetry, Node Exporter, Vector, Logstash, Fluentbit and more
Before, our LLM business had a slow, unstable logging system on Loki. Now, Apache Doris manages massive logs, ensuring over 99.9% availability and sub-second query latency on 100 million logs.
At Netease, high storage costs and slow queries on Elastcisearch were issues. Now, Apache Doris has cut storage costs by 70%, let us use SSDs for hot data at no extra cost, and sped up queries 11x with less CPU use.
We generate 14 billion log entries daily, totaling 80TB, with a total archive exceeding 40PB. Our old platform based on Elasticsearch suffered from high storage costs and poor ingestion performance... Switching to Apache Doris cut costs by 50% and sped up queries by 2 to 4 times.
Low Cost, High Performance, Open AI Observe Stack
AI Observability
Automatically instrument AI SDKs including Langfuse, Langchain, OpenAI, Claude and more. Track exactly every step of AI agent and LLM to help you understand and improve your application.
Infra Observability
Collect logs, traces, metrics using OpenTelemetry SDKs, Vector, Logstash and more. Analyze logs, traces, and build dashboards, alert in Grafana to identify system issues and increase stability.
More Featured Resources
1. Observability based on VeloDB
2. Elasticsearch vs ClickHouse vs Apache Doris — which powers observability better?
3. Why Apache Doris is a Better Alternative to Elasticsearch for Real-Time Analytics
4. Apache Doris for log and time series data analysis in NetEase, why not Elasticsearch and InfluxDB?
5. Creator of Talkie migrated from Loki and built a PB-scale logging system with Apache Doris
6. What is Observability
7. Demo: Building an Observability Platform with OpenTelemetry, VeloDB, and Grafana
8. Collect, store, and visualize logs using VeloDB Cloud
9. Benchmark: Elasticsearch vs. VeloDB Cloud for log analytics
10. Trace data analytics: OpenTelemetry and VeloDB
