Joined May 2022
- Tweets 342
- Following 1,109
- Followers 377
- Likes 209
239 Photos and videos
Apache Doris handles agent observability data with a hybrid data modeling in JSON storage:
1️⃣ Regular columns for stable, high-frequency fields
2️⃣ VARIANT for dynamic JSON payload
3️⃣ Inverted indexes on VARIANT fields and long text
🔗velodb.io/blog/json-in-agent…
48
Apache Doris 4.1 introduces per-user identity mode with Polaris.
Doris now forwards each user's real identity to the open catalog instead of using a shared account.
Demo: velodb.io/blog/apache-doris-…
1
25
This Thursday in Sydney, come meet us and @dataengbytes at Stone & Chalk for an evening on real-time analytics for AI systems.
Apache Doris PMC Mingyu Chen will discuss why real-time analytics is becoming essential in the AI #agent era.
👉 RSVP: velodb.io/events/sydney-meet…
1
1
59
Join us in San Francisco on June 23 for an evening with @redpandadata and @datastrato on building the real-time data stack for AI agents.
- Streaming ingestion
- Metadata Management
- Real-time analytics and hybrid search
👉 Register: luma.com/agc8cwc1?utm_source…
#agents
1
1
2
138
#Iceberg is evolving from just a table format to a key data foundation for real-time analytics, lakehouse, and AI.
Join us for a webinar on running Iceberg in real production scenario. Lessons from Lucid Motors.
👉RSVP: velodb.io/events/lucid-velod…
67
Apache Doris 4.1 adds Iceberg V3 support:
- Read
- Write (including UPDATE, DELETE, MERGE INTO)
- DDL
- Table maintenance
- Diagnose
This update Doris can now covers the full Iceberg operational lifecycle.
🔗velodb.io/blog/apache-doris-…
#Iceberg
35
Come meet us at PGday Boston! Let's chat about PG OLAP architecture.
We offer native CDC sync from Postgres WAL, real time, no middleware to manage.
Postgres stays as the transactional source of truth, and Doris can handles the analytical side.
🔗2026.pgdayboston.org/registr…
31
We built AgentLogsBench for AI agent observability
The benchmark test how databases handle four core access patterns of AI agent observability
Combined score:
Doris: 1.28x
Elasticsearch: 3.11x
ClickHouse: 11.91x
DuckDB: 73.48x
Postgres: 439.55x
🔗velodb.io/blog/agentlogsbenc…
108
Running a RAG demo is very different from deploying a retrieval system for production
Apache Doris has helped firms like ByteDance to build production-ready retrieval system
Three key decisions:
- Chunk Shape
- Embedding Strategy
- Vector Index
🔗velodb.io/blog/the-chunking-…
1
38
We have upgraded the Spill to Disk feature in Apache Doris 4.1, addressing OOM (out of memory) issues
1️⃣ Full operator coverage: Hash Join, Aggregation, and Sort operators all support Spill to Disk.
2️⃣ Recursive spill
3️⃣ Dynamic triggers
🔗velodb.io/blog/apache-doris-…
34
VeloDB (Powered by Apache Doris) retweeted
Sydney data engineers 👋
Special meetup with @VeloDB_IO — Thu 11 June, 5:30pm @ Stone & Chalk, Tech Central
Two talks on real-time analytics for AI systems live demo
Free. Food provided. Great people.
RSVP 👇
1
1
1
59
Testing Apache Doris, ClickHouse, Elasticsearch, and PostgreSQL in wide #JSON workloads
Setup:
- 10K JSON paths, no fixed schema
- Each row fills 100 random keys (1% sparsity)
- 100M rows, 160 GB, ingested across 1,000 files
- 16 cores, 64 GB RAM, SSD
🔗velodb.io/blog/beyond-10k-fi…
1
2
124
Last Call 📣: This May 18-19, the VeloDB team'll be at AI & Big Data Expo North America.
If your team is evaluating AI data infra or looking to consolidate search and analytics stack, find us at Booth 252.
📍 San Jose McEnery Convention Center, CA
🎟️ai-expo.net/northamerica/
46
Come join us in San Francisco on May 20 to build a data layer that can keep up with the real-time, high-concurrent, and multimodal demands of enterprise agents.
Partner @puppyquery and @cocoindex_io
📅 Wed, May 20, 5:30 – 7:30 PM
📍 Trellis Cafe, SF
🎟️ lu.ma/5zskcl5y
1
35
VeloDB supports lakehouse analytics natively, querying and write data for Iceberg table (V2/V3) directly on object storage, without moving data.
See how VeloDB performs in lakeshoue queries:
🔗 Full benchmarks technical breakdown: velodb.io/blog/velodb-perfor…
1
1
50
Come join us in Bengaluru this Saturday for Lakehouse at Scale: Apache Doris × OLake @_olake
- Apache Doris powers AI agents with hybrid search
- Iceberg table maintenance at CDC scale
📅Saturday, May 9, 11:00 AM
📍 Hustlehub Tech Park, Bengaluru
🎟️ luma.com/s91b12i5
1
49
Check out @khameshra's demo for setting up an Iceberg lakehouse with Apache Doris in 15 minutes.
The stack:
PostgreSQL → @_olake (CDC) → Apache Iceberg on Object Storage → Apache Doris (Query Engine)
🔗 Full walkthrough: velodb.io/blog/set-up-a-lake…
39
Apache Doris 4.1 is here
🚀New Feature Highlights:
1. New IVF and IVF_ON_DISK indexes for Vector Search
2. Better Full-Text Search: BM25 scoring
3. Supporting Long-Context #AI: 100MB JSON storage
4. Unified Lakehouse: Iceberg V2/V3 read and write
🔗velodb.io/blog/apache-doris-…
52
Join us and Supermetal for a webinar to build a real-time analytics stack with lean, affordable real-time CDC.
👉 RSVP: velodb.io/events/supermetal-…
Supermetal offers real-time CDC as a single Rust binary. Pair it with VeloDB to get continuous ingestion and sub-second queries.
1
1
44
We built a Snowflake alternative on #AWS
Results for a team that migrated from Snowflake to the alternative:
- 80% cost reduction
- Ingestion latency down 5-10 minutes to 1-2 seconds
- 90x query performance boost on 475M row scans: 90x improvement
🔗velodb.io/blog/real-time-ana…
58