How AnalyticDB Powers Double 11: Cloud‑Native Data Warehouse Innovations

AnalyticDB is a cloud‑native data warehouse that fully supports the MySQL protocol and SQL:2003, enabling real‑time multi‑dimensional analysis on massive data.

Double 11 performance

During the 2020 Double 11 shopping festival AnalyticDB handled over 600 billion orders, achieving a write TPS peak of 2.14 × 10⁸, processing 174 571 online ETL jobs per second and 7.7 trillion rows in real time.

Challenges

Key challenges include entering core transaction links with peak 5 million TPS, supporting high‑concurrency write and query, and integrating with production workflows such as Cainiao warehousing and data‑bank services.

Key technologies

Compute‑storage separation : A new shared‑nothing architecture provides a serverless storage layer and independently elastic compute, allowing scaling without data migration.

Hot‑cold storage tiering : Tables can be defined as HOT, COLD or MIXED; hot partitions reside on SSD, cold on OSS, with automatic migration driven by background Build tasks.

Cold‑data query acceleration : SSD cache, multi‑granularity cache entries, metadata pre‑warming and lock‑free queues improve cold‑partition access.

Resource groups (multi‑tenant) : Isolated compute pools support interactive or batch query modes, with elastic node counts and per‑group configuration.

Elastic scaling : Time‑based elasticity automatically expands resources before traffic spikes, reducing cost during low‑load periods.

Intelligent optimization : Real‑time monitoring adjusts execution plans, statistics and engine parameters; dynamic plan tuning and materialized view management further boost performance.

SQL examples

create table event(
  id bigint auto_increment,
  dt datetime,
  event varchar,
  goods varchar,
  package int,
  ...
) distribute by hash(id)
partition by value(date_format(dt, '%Y%m%d')) lifecycle 365
storage_policy = 'MIXED' hot_partition_count = 7;

Resource‑group commands

-- Create resource group
CREATE RESOURCE GROUP group_name
    [QUERY_TYPE = {interactive, batch}]
    [NODE_NUM = N];

-- Bind resource group
ALTER RESOURCE GROUP BATCH_RG ADD_USER=batch_user;

-- Adjust size
ALTER RESOURCE GROUP BATCH_RG NODE_NUM=10;

-- Drop resource group
DROP RESOURCE GROUP BATCH_RG;