How Alibaba’s Database Tech Evolved for Double 11: 10 Years of Innovation

As the tenth Double 11 approaches, Alibaba Technology launched the "Ten Years of Code" series, inviting core engineers who have prepared for Double 11 over the years to review the evolution of Alibaba’s technology.

Researcher Zhang Rui from Alibaba’s Database Division shares the untold stories of database technology behind Double 11, illustrating how each increase in zero‑hour transaction volume was driven by breakthroughs in database engineering and the relentless spirit of Alibaba technologists.

Over the past ten years, Alibaba’s technology role shifted from using Double 11 to drive tech development to creating new commerce through technology, with many solutions now offered as cloud services to various industries.

Three development phases are identified:

2005‑2009: Commercial database era (IOE era)

2010‑2015: Open‑source database era (AliSQL, TDDL, SSD, X86)

2016‑present: Self‑developed database era (X‑DB)

The "去IOE" transition replaced Oracle, IBM mainframes, and EMC storage with distributed middleware TDDL, AliSQL (Alibaba’s MySQL branch), high‑performance X86 servers and SSDs, delivering three major benefits: horizontal scalability, autonomous feature development, and improved stability through multi‑active, cross‑city architectures.

In 2016 Alibaba began developing its own native distributed database, code‑named X‑DB, to meet three needs: a globally deployable database similar to Google Spanner, the extreme performance demands of Double 11, and the rising storage cost caused by hot‑cold data patterns that could no longer be satisfied by incremental improvements to open‑source databases.

To reduce machine‑resource costs, Alibaba first adopted high‑performance ECS instances and hybrid‑cloud VPC networking for Double 11, then introduced online‑offline mixed deployment, achieving storage‑compute separation that allowed the database to use offline compute resources during peak traffic.

Intelligent database features also emerged: self‑driving capabilities such as automatic SQL optimization, space optimization, and the CloudDBA diagnostic engine, built on full‑SQL capture via AliSQL and a time‑series TSDB for second‑level monitoring.

Key milestones (2012‑2018) :

2012: First Double 11 experience; manual QPS estimation, capacity overload, and severe over‑selling.

2013: Inventory hotspot optimization and the introduction of "shadow tables" for full‑link testing, enabling realistic online pressure testing.

2015: Media‑screen GMV display challenge solved with a 3‑second latency for the first real‑time number.

2016: Monitoring redesign using TSDB and full‑SQL capture; launch of CloudDBA.

2016‑2017: Integration of Alibaba’s cache Tair with X‑KV and TDDL to provide high‑performance KV interfaces alongside SQL.

2016‑2017: Pre‑warming system to load hot data and JIT compilation before zero‑hour, smoothing transaction curves.

2017‑2018: Storage‑compute separation built on Ceph and Alibaba’s Pangu 2.0, extensive use of 25 Gbps TCP and RDMA, and DBFS zero‑copy I/O, achieving near‑local latency.

These innovations transformed Alibaba’s database platform from a commercial‑database‑dependent system to a self‑sufficient, cloud‑native, intelligent service that now powers not only Double 11 but also millions of customers worldwide.

In summary, the six‑year span from 2012 to 2018 witnessed continuous transaction‑volume growth, repeated database breakthroughs, and an unwavering commitment to turning the "impossible" into the "possible".

Looking forward, Alibaba anticipates another decade of even greater advancements.