Solution
Double 11 Global Shopping Festival operates 100% cloud-native infrastructure, processing 583,000 peak transactions per second during the world's largest online shopping event. Apsara computational engine orchestrates elastic scaling to handle 1,457x growth from the inaugural 2009 event while Hanguang 800 AI inference chip accelerates product search and recommendation engines. M6 large-scale AI model optimizes apparel design workflows, reducing design cycles from months to weeks. The cloud-native architecture reduced computing resources by 50% per 10,000 transactions, boosted CPU utilization by 30%, and enabled green technologies including liquid cooling and renewable energy reducing carbon emissions by 26,000 tons annually.
Data flow
Customer transactions → Apsara engine for elastic scaling → recommendation/search via Hanguang 800 → order processing → last-mile logistics assignment → fulfillment
Solution architecture
5 components · 1 layer - Compute
- Elastic Compute Service Foundational compute resources handling peak transaction processing and traffic distribution
- Apsara Computational Engine Super computational backbone orchestrating elastic scaling for unpredictable traffic surges during Double 11
- Machine Learning Platform for AI ML algorithms for recommendation systems, fraud detection, and user profiling
- Hanguang 800 AI Inference Chip Specialized AI chip accelerating product search, recommendation, and personalization efficiency
- M6 Large-Scale AI Model Foundation model optimizing AI-intensive operations including apparel design and product analytics
- 100% cloud-native applications on Alibaba Cloud
- Apsara super computational engine for elastic scale
- Hanguang 800 AI inference chip for search and recommendation
- Liquid cooling + wind energy in hyperscale data centers
- M6 large-scale AI model for AI-intensive operations
The algorithm performance enhanced the search function by 200%, with a 58% reduction in energy costs.
The event in 2020 saw orders peaking at 583,000 transactions per second, 1,457 times that of the first Double 11 in 2009, exceeding the highest transactions of 544,000 of 2019.
The full use of cloud-native technologies reduced computing resources by 50% for every 10,000 transactions compared to last year.
