In high‑concurrency systems, tail latency is driven primarily by waiting on locks, resources, and scheduling rather than raw computation, with phenomena like head‑of‑line blocking, context‑switch overhead, and cache‑coherency costs amplifying unpredictable delays.
This article explores the fundamental architecture of a payment platform, detailing the separation of transaction and payment cores, the key modules such as transaction abstraction, payment orchestration, service governance, data consistency, asynchronous processing, performance testing, and practical strategies for stability and scalability.
This article explains how a typical payment platform is structured into transaction and payment cores, details the interactions between services, describes governance mechanisms such as unified context and data consistency, and shares practical production techniques like performance testing, async processing, and service degradation.
This article walks through the complete architecture of a payment platform, detailing the transaction and payment cores, their interactions, service governance, data consistency, DB sharding, asynchronous processing, performance testing, and practical production practices for building robust backend payment systems.
A recent CCF‑A conference paper reveals that an adaptive K‑value backoff lock, derived from queueing theory and implemented in Apache RocketMQ, can replace both spin and mutex locks, achieving up to 37.58% performance gains on x86 CPUs and 32.82% on ARM while reducing CPU usage and resource consumption.
Facing a traffic surge that pushed a risk‑control engine's response time beyond 250 ms, the team applied a series of systematic optimizations—feature prefetching, batch requests, asynchronous accumulator updates, multi‑level caching with Bloom filters, and log‑compression redesign—resulting in latency dropping to 25 ms, CPU and memory usage falling by up to 90%, and storage costs reduced by over a third.
This article details the architectural evolution of JD's VOP message warehouse, describing the challenges of massive data volumes, the transition from V1.0 to V3.0+ through database sharding, MongoDB adoption, traffic governance, stability improvements, and cost reduction strategies, while presenting performance metrics and future outlook.
This article walks readers through the complete point‑to‑point, end‑to‑end payment chain in China, using a JD.com purchase of Three Squirrels nuts as a concrete example to illustrate the roles of the cashier, order, transaction, payment, settlement, and accounting subsystems within a typical e‑commerce payment architecture.
This article revisits the Redlock debate by examining Antirez's rebuttal to Martin Kleppmann, analyzing Hacker News discussions, comparing Redis, ZooKeeper, and Chubby lock implementations, and exploring timing assumptions, clock skew, and fencing token mechanisms to assess distributed lock safety.
