Ant Financial’s Financial‑Grade Cloud Native Architecture and Service Mesh Deployment for Double 11 2019
The article details Ant Financial’s large‑scale, financial‑grade cloud‑native architecture and its Service Mesh implementation during the 2019 Double 11 shopping festival, describing the evolution from monolithic to distributed systems, the challenges of mesh adoption, and the operational benefits achieved.
On November 19, 2019, Ant Financial held a technology summit in Beijing to showcase the technical foundations behind the 2019 Double 11 Alipay peak, announcing the new OceanBase 2.2 release and the SOFAStack dual‑mode micro‑service platform.
Yang Bing, General Manager of Ant Financial’s FinTech Product Technology Department, presented a full transcript of his talk on the massive application of Ant Financial’s financial‑grade cloud‑native architecture during Double 11.
The 2018 Double 11 marked the first use of a cloud‑native architecture by Ant Financial and MyBank; by 2019 the Service Mesh architecture covered 100% of the core payment chain, becoming the industry’s largest Service Mesh cluster.
Transaction peaks have grown year over year, driving architectural evolution from a centralized system to a distributed, cloud‑native platform that can reliably support business growth.
While architectural upgrades solve many business pain points, they also impose heavy migration and upgrade burdens on development teams, prompting the desire for a platform that abstracts infrastructure concerns away from business code.
The first stage of financial transaction technology evolution involved building a financial‑grade distributed architecture based on SOFAStack and OceanBase, emphasizing end‑to‑end consistency, high availability, scalability, and performance.
OceanBase provides distributed transaction support at the database layer, but cross‑cluster consistency must be handled by middleware, illustrating the need for comprehensive end‑to‑end guarantees.
Legacy systems and heterogeneous environments create coupling challenges; Service Mesh offers an additional abstraction layer that decouples business logic from infrastructure, enabling independent evolution.
Conway’s Law is addressed by Service Mesh, aligning software architecture with organizational structure.
Ant Financial’s SOFAMesh was deployed at Double 11, achieving 100% coverage of the core payment link, scaling to dozens of thousands of containers, handling tens of millions of QPS with minimal performance overhead, and enabling rapid iteration (10+ releases per month).
Traditional sidecar injection incurs high resource costs and operational risk; Ant Financial adopted an “injection upgrade” approach, sharing CPU and memory between the sidecar and the application to minimize overhead.
For sidecar upgrades, Ant Financial implemented a hot‑migration strategy that transfers traffic seamlessly between twin pods, ensuring zero‑impact upgrades and preserving service stability.
During the Double 11 promotion, precise traffic control was required to switch between active and standby service groups within a narrow time window, achieved through fine‑grained traffic steering, keep‑alive traffic, and Service Mesh‑driven flow redistribution.
Service Mesh provides global traffic management capabilities such as multi‑version releases, load‑testing traffic diversion, and service grouping, enabling fine‑grained control even in complex call graphs.
Ant Financial is open‑sourcing SOFAMesh, which can be deployed on cloud‑native platforms or virtual machines, supports multiple protocols, and integrates with SOFAStack, Dubbo, and Spring Cloud ecosystems.
Future work includes unified data‑center traffic management across north‑south and east‑west flows, and broader registration‑center compatibility.
AntTech
Technology is the core driver of Ant's future creation.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.