0 views collected around this technical thread.
This article explains the principles, advantages, and drawbacks of CAS, LongAdder, and AtomicLong in Java, illustrating why Alibaba recommends LongAdder for high‑concurrency counter scenarios, how the ABA problem arises, and how AtomicStampedReference can mitigate it.
This article analyzes how Tencent Channel builds a scalable Feeds system for a massive multi‑level community by presenting a three‑layer architecture, core design principles, and read/write diffusion strategies to handle sub‑channel lists, post squares, and personal feeds while mitigating empty‑pull and moderation challenges.
This article analyzes the overselling problem that occurs during high‑concurrency flash‑sale (seckill) operations and presents seven concrete solutions—including improved lock placement, AOP‑based locking, three types of database locks, optimistic locking, a blocking‑queue approach, and a Disruptor queue—complete with SpringBoot 2.5.7 code samples, performance test results, and practical recommendations.
The article explains Flink’s post‑submission scheduling pipeline—from Dispatcher creating SchedulerNG and building the ExecutionGraph, through pipelined region construction and the PipelinedRegionSchedulingStrategy, to slot sharing allocation—identifying why slot and TaskManager overloads occur and proposing randomization or fine‑grained resource strategies to balance load.
The article uses a safety‑principle analogy to introduce three real‑world backend incidents—a Java ternary‑operator NullPointerException caused by unintended unboxing, a RocketMQ message‑storm triggered by swapped parameters, and a network‑path latency spike—showing how language semantics, messaging contracts, and infrastructure topology each can create production bugs.
This article analyzes why simple @Transactional and lock annotations still cause overselling in flash‑sale scenarios, then presents seven backend solutions—including improved controller locking, AOP locking, pessimistic and optimistic database locks, and queue‑based approaches with BlockingQueue and Disruptor—along with code samples and JMeter test results.
This article explains why RocketMQ guarantees at‑least‑once delivery, describes the three typical duplication scenarios, evaluates simple idempotent approaches, discusses transactional and non‑transactional deduplication solutions—including a Redis‑based method—and outlines their advantages, limitations, and practical recommendations for achieving exactly‑once processing in distributed systems.
This article introduces the new full‑process features of OpenMLDB 0.4.0, explains its unified online/offline storage, high‑availability task management, and end‑to‑end AI workflow, and provides step‑by‑step instructions for quickly deploying both single‑node and cluster versions to run a complete online AI application.
This article explains how Hystrix implements resiliency patterns such as bulkhead, circuit breaker, retry, and degradation for microservice calls, details its execution workflow, core components, dynamic configuration, isolation strategies, metrics collection, and practical usage, and discusses future alternatives and extensions.
This article explains how JD.com’s JUST platform leverages the open‑source columnar database ClickHouse to store, query and analyze massive time‑series data, covering data modeling, lifecycle management, system goals, technology selection, cluster architecture, deployment, scaling and future enhancements.
This article details the design principles, implementation steps, and code examples for a lightweight transactional message module in a microservices environment, leveraging Spring Boot, RabbitMQ, MySQL, and Redisson to achieve low‑intrusion, eventual‑consistency messaging with compensation and retry mechanisms.