Insights from Leading Architects on System Refactoring, Platform Evolution, and High‑Performance Architecture
The article compiles expert experiences from major Chinese tech firms on progressive system refactoring, platform‑centric redesign, service‑oriented evolution, and high‑performance architecture, highlighting practical steps, technology choices, and lessons learned for modern backend and operations engineering.
Industry architects from aerospace, search, food delivery, smartphone, travel, ride‑hailing, classifieds, and instant‑messaging companies share their practical approaches to modernizing legacy systems, balancing continuous business delivery with architectural improvement.
Progressive Refactoring (航天信息)
Emphasizes gradual decoupling of business domain layers from technical platforms, parallel optimization and maintenance, and simultaneous platform and reconstruction efforts, culminating in a seven‑step software refactoring process.
Sogou Commercial Platform Evolution
Describes four stages—initial monolith, horizontal scaling, service‑oriented architecture, and stream processing—highlighting database sharding with the Compass framework, stateless computation, and a Kafka‑Storm‑Pump pipeline.
Ele.me Overall Architecture
Shows evolution from single‑machine to cluster to SOA, using F5 for gateway, domain‑based database sharding, hot‑data caching, async messaging, service degradation, a custom DAL middleware, and a Python‑based core service.
Mi.com Architecture Transformation
Details DB partitioning, star‑topology with MQ for decoupling, three‑tier design, Cobar for traffic spikes, and a Thrift‑based RPC framework with etcd for service discovery and multi‑language support.
Tuniu Mobile Architecture
Focuses on service‑oriented migration, multi‑data‑center deployment, performance tuning with Codis caching, custom BWT cache updater, OSS monitoring, and plugin‑based mobile client development.
Kuaidi (Didi) Architecture Practice
Highlights distributed transformation using Dubbo + RocketMQ, log collection with Log4j‑Flume‑Elasticsearch, real‑time analytics via RocketMQ‑Storm‑HBase, and a mock MySQL‑slave pipeline feeding HBase for low‑latency queries.
58.com Mobile Push Platform
Tracks push service from iOS‑only to multi‑platform, introducing a unified push platform, parallelized/asynchronous delivery, and device‑specific channel selection for stability.
Tencent AMS Architecture
Explains a three‑layer design (frontend, CGI, service), NoSQL for CRUD performance, optimistic CAS locking for high‑concurrency, horizontal scaling, failover, overload protection, and a comprehensive security and monitoring framework.
Overall Takeaways
Key themes include full‑stack architectural responsibility, convergence on service‑orientation, sharding, async messaging, and monitoring, while also recognizing divergent needs in security and consistency; architects must continuously embrace change driven by business growth and emerging technologies.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
IT Architects Alliance
Discussion and exchange on system, internet, large‑scale distributed, high‑availability, and high‑performance architectures, as well as big data, machine learning, AI, and architecture adjustments with internet technologies. Includes real‑world large‑scale architecture case studies. Open to architects who have ideas and enjoy sharing.
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.