1. 演讲人：杨勇强

2. AI 时代的实时性、全球化、成本与复杂性鸿沟 AI 时代实时数据分析进化 03 AI 实时数据分析实践案例

3.

4. 01 AI 时代的实时性、全球化、成本与复杂性鸿沟

5. Apache Doris：Open-Source Real-Time Analytics and Search Database 20 13 Project Creation 13k+ GitH ub Stars #1 Daily Page View s of Apache Project Websites 20 17 Open Source 640+ Contributors 2022 ASF Top Project 500 0+ #1 Monthly Activ e C ontributo rs Enterprises

6. AI 时代的实时性、全球化、成本与复杂性鸿沟 AI 对数据实时性和查询速度要求越来越高 LLM 带来新的数据安全和合规风险 海量数据存储分析成本高昂 数据技术栈臃肿，维护复杂性高

7. 02 AI 时代实时数据分析进化

8. AI 时代数据分析进化 AI-boosted AI-driven existing analytical workloads new analytical workloads

9. AI-boosted existing analytical workload Faster, more real-time, more data, more unstructured text Real-Time Analytics Data Warehousing Observability

10. Real-Time Analytics From User-Facing Analytics to Agent-Facing Analytics, more real-time and faster User-Facing Analytics Agent-Facing Analytics As cloud computing and SaaS software become more popular, With the rise of AI technologies, especially AI Agents, more embedding analytics into applications is crucial. This is also called analytical decisions will be made automatically by AI. This will customer-facing or user-facing analytics. improve efficiency and accuracy in decision-making. Order Analytics ｜ Advertising Analytics ｜ Inventory Analytics Fraud Detection｜ Ad Serving｜ Personalized Recommendation Agent-Facing Analytics - Real-Time Ingestion & Update: ~ 1s minimum data latency Why Choose Doris - Blazing-Fast Analytics: < 100ms average query latency - High-Concurrent Queries: - Agent Native: Doris > 10,000QPS maximum query concurrency MCP Server

11. Data Warehousing From triditional warehouse to Lakehouse + Real-Time Analytics Database Open Data Lakehouse Lakehouse is designed for both AI and analytics workload Why Choose Doris (Doris) Build the Fastest Real-Time Data Warehouse in the Lakehouse， replacing Trino/Presto，SparkSQL，... - Fast: 3x to Trino/Presto - Open - Unified

12. Observability From Cloud Native + Microservice to LLM + AI Agent, more data and text 2013: The concept of observability began to gain traction as companies like Twitter started to adopt it to manage History of Observability their increasingly complex distributed systems. 2020: Observability became a hot topic in DevOps, with more companies recognizing its importance in managing complex IT architectures Metrics Traces - 10x Cost Effective Compared to Elasticsearch Why Choose Doris - Flexible Semi-Structured Data Variant Type - Open Integration for ELK, OpenTelemetry, Grafana and more Logs

13. AI-driven New Analytical Workload AI-boosted AI-driven existing analytical workloads new analytical workloads

14. Hybrid Search Why Vector-Capable General-Purpose Databases are Better for Enterprise GenAI ？ - Lack of Hybrid Query Capabilities - Limited Integration with Structured Data - Operational Complexity and Increased Costs - Large scale data volume, eg PBs

15. AI Operators in SQL Integrating GenAI for Enhanced Text Analysis in Database LLM_SUMMARIZE LLM_TRANSLATE LLM_SENTIMENT LLM_CLASSIFY LLM_GENERATE LLM_EXTRACT LLM_FILTER LLM_SIMILARITY LLM_FIXGRAMMAR LLM_MASK

16. 03 实践案例

17. “基于 SelectDB 的新系统已接入 MiniMax 内部所有业务线日志数据，满足 10GB/s 高吞吐实时写入和秒级查询的需求 通过存算分离比自建 Doris 计算资源降低 40%，热数据存储资源降低 50%”

18. “腾讯音乐内容库基于 Doris 替换 ES 和 CK，同时满足搜索和分析的需求，存储成本降低 80%，写入性能提升 4倍， 增强大模型与 OLAP 交互效率、结果输出的准确率，最终提供更智能化的问答交互服务”

19. “ByteDance 基于 Doris 开发了向量索引和全文检索相关性打分，向量检索性能提升10倍 支持 HybridSearch，满足多个业务 RAG 和 GraphRAG 需求”

20.

21.