Author

WeChat Backend Team

Official account of the WeChat backend development team, sharing their experience in large-scale distributed system development.

Articles

Likes

Views

Comments

Latest from WeChat Backend Team

47 recent articles

WeChat Backend Team

Oct 23, 2024 · Artificial Intelligence

How We Scaled AI Computing in WeChat with Ray: From Challenges to AstraRay

This article details the AI computing challenges faced by WeChat, explains why the Ray distributed engine was chosen, and describes the design and large‑scale deployment of the AstraRay platform—including scheduling, resource management, and multi‑model support—to achieve low‑cost, high‑efficiency AI services.

AI PlatformAstraRayDistributed Computing

0 likes · 20 min read

How We Scaled AI Computing in WeChat with Ray: From Challenges to AstraRay

WeChat Backend Team

Oct 25, 2023 · Fundamentals

Mastering Metric Covariance for Accurate A/B Test Analysis

This article explains the statistical foundations of A/B testing, introduces potential outcomes and average treatment effect, defines metric covariance, and presents practical estimation methods—including naive, data‑augmentation, and bucket‑based approaches—along with real‑world performance evaluations and applications such as variance reduction and Bayesian optimization.

A/B testingbayesian optimizationexperimental design

0 likes · 18 min read

Mastering Metric Covariance for Accurate A/B Test Analysis

WeChat Backend Team

Jun 13, 2023 · Artificial Intelligence

Boosting Vertical Federated Learning: Optimizing Paillier Encryption & Model Stability

This article examines the challenges of data privacy in big‑data environments and presents a comprehensive approach to vertical federated learning, detailing framework optimizations, Paillier homomorphic encryption enhancements, PSI‑based feature selection, and adversarial learning techniques to improve model stability and deployment on a unified ML platform.

Paillier encryptionPrivacy Computingfederated learning

0 likes · 19 min read

Boosting Vertical Federated Learning: Optimizing Paillier Encryption & Model Stability

WeChat Backend Team

Jun 7, 2023 · Artificial Intelligence

How TransE+ Boosts Knowledge Graph Embedding on WeChat’s Plato Framework

This article presents the development and deployment of the TransE+ knowledge‑graph embedding model on the Plato graph‑computing platform, detailing its architectural upgrades, training optimizations, performance gains, and business‑oriented adaptations for large‑scale real‑world applications.

AIEmbeddingTransE+

0 likes · 22 min read

How TransE+ Boosts Knowledge Graph Embedding on WeChat’s Plato Framework

WeChat Backend Team

Jun 1, 2023 · Big Data

How WeChat Boosted Flink Stability with TaskManager Recovery and Load Balancing

This article details WeChat’s Gemini‑2.0 real‑time streaming platform built on Flink, explaining two key stability enhancements: a TaskManager‑level partial failure recovery that avoids data loss during node crashes, and a load‑balancing scheduler that evenly distributes tasks across TaskManagers to improve resource utilization and reduce latency.

Big DataFlinkKubernetes

0 likes · 16 min read

How WeChat Boosted Flink Stability with TaskManager Recovery and Load Balancing

WeChat Backend Team

May 24, 2023 · Databases

Boost ClickHouse Bitmap Queries 10x with BitBooster: Techniques & Results

This article explains how the BitBooster suite accelerates ClickHouse bitmap (BitMap) queries by up to tenfold, covering background, performance bottlenecks, single‑node and read optimizations, layout and instruction‑set enhancements, encoding dictionaries, multi‑node scaling, and real‑world benchmark results.

BitmapClickHouseOptimization

0 likes · 23 min read

Boost ClickHouse Bitmap Queries 10x with BitBooster: Techniques & Results

WeChat Backend Team

May 17, 2023 · Big Data

Boosting Real-Time Recommendations: Apache Pulsar Optimizations at WeChat

This article details how WeChat's Gemini‑2.0 big‑data platform leverages Apache Pulsar, outlining cloud‑native advantages, load‑balancing refinements, cache and SSD tuning, high‑availability safeguards, and cost‑saving strategies that together enable large‑scale, real‑time, deep‑learning recommendation workloads.

Apache PulsarBig DataMessage Queue

0 likes · 17 min read

Boosting Real-Time Recommendations: Apache Pulsar Optimizations at WeChat

WeChat Backend Team

Aug 5, 2022 · Artificial Intelligence

How WeChat’s Ekko Achieves Ultra‑Low‑Latency Model Updates for Billion‑User Recommendations

At the 16th OSDI conference, Tencent’s WeChat team presented the award‑winning Ekko system—a groundbreaking, ultra‑low‑latency model‑update solution for massive recommendation workloads that dramatically speeds up updates, supports over a trillion‑scale models, and has already boosted user engagement across billions of daily users.

Model UpdateWeChatlarge scale

0 likes · 5 min read

How WeChat’s Ekko Achieves Ultra‑Low‑Latency Model Updates for Billion‑User Recommendations

WeChat Backend Team

Jun 7, 2021 · Artificial Intelligence

How WeChat’s TFCC Boosts Deep Learning Inference Performance Across Platforms

The TFCC framework, developed by WeChat's backend team, delivers high‑performance, easy‑to‑use, and universal deep‑learning inference by supporting numerous ONNX and TensorFlow operations, optimizing model structures, constants, and operators, and providing a versatile runtime and math library for both CPU and GPU platforms.

InferenceONNXTFCC

0 likes · 8 min read

How WeChat’s TFCC Boosts Deep Learning Inference Performance Across Platforms

WeChat Backend Team

Mar 6, 2021 · Backend Development

How We Scaled a Live Chatroom to 15 Million Concurrent Users

This article details the evolution of a WeChat live‑room chat component from its 1.0 high‑performance design to a 2.0 architecture that overcomes scalability, reliability, and traffic‑isolation challenges, enabling a single room to support up to 15 million simultaneous online users.

ChatroomHyperLogLogScalability

0 likes · 13 min read

How We Scaled a Live Chatroom to 15 Million Concurrent Users