Why databases are key to Alibaba’s Singles’ Day sales

At the stroke of midnight on this year’s Singles’ Day, online shoppers in China and across Asia started placing orders on Alibaba’s e-commerce sites that saw peak orders hit a whopping 583,000 per second.

This was 1,400 times the peak volume of the inaugural event 12 years ago on 11 November 2009. To cope with the record number of orders, which amounted to $74.1bn in gross merchandise value this year, Alibaba had to shore up its infrastructure.

With each order split into multiple database transactions, ensuring the scalability and elasticity of Alibaba’s database systems was key, said Li Feifei, president of database business at Alibaba Cloud Intelligence in an interview with Computer Weekly.

By using cloud-native databases that support online transaction processing, as well as offline batch processing and big data workloads, Alibaba had been able to achieve zero downtime in what has now become the world’s largest online shopping event.

As an example of the workloads that Alibaba’s database systems were grappling with this year, PolarDB, a relational database, set a new record with 140 million queries per second during peak time, a 60% increase from last year. In addition, AnalyticDB, the company’s self-developed cloud-native data warehouse, processed up to 7.7 trillion lines of real-time data, equalling 15 times the data contained in the UK web archive at the British Library.

The same database systems were also used by China Post to deal with over 100 million orders during the shopping event, with about 100,000 China Post users checking their parcels’ real-time status online.

Making its commercial debut this year was Lindorm, Alibaba’s cloud-native multi-model database that is compatible with multiple open-source interfaces such as HBase, Phoenix, OpenTSDB and Solr.

Lindorm offers capabilities such as SQL query, aggregation computing, retrieval analysis, and serverless elasticity, to enable the storage and analysis of structured, semi-structured and unstructured data, as well as shared storage for multi-model data.

As the core basic storage of Alibaba, Lindorm was designed based on a high-availability architecture to eliminate a single point of failure and has successfully supported Singles Day sales for over a decade.

“Alibaba Cloud has been at the forefront of cloud database product offerings for many years,” Li said. “In the cloud-native era, it is even more pivotal for businesses to be able to take advantage of innovative database-as-a-service [DBaaS] offerings to enhance and support high-concurrency and high-volume web applications such as e-commerce, online gaming and financial technology.

“Many Fortune 500 firms are already using our Cloud DBaaS for mobile apps, backups and tests, and the most recent impressive performance of our proprietary solutions during the 11.11 global shopping festival is yet another strong endorsement on our capability.”

The high volume of database workloads did not require the use of more compute resources. Not only did Alibaba Cloud support one of the world’s largest container clusters, upscaling to as many as one million containers in an hour, it was able to reduce the amount of computing resources needed for every 10,000 transactions by 80% compared to four years ago.

“We were very proud to support 800 million consumers and 250,000 brands during the world’s largest shopping festival,” said Li Cheng, chief technology officer of Alibaba Group.

“From the robust digital infrastructure supporting zero downtime operation, to cloud-native offerings for developers’ efficiencies and consumer-facing applications for creating some of smoothest engagement experiences, Alibaba’s technologies have once again passed the toughest tests with flying colours.”