Comparison

So Sánh Alipay Stack với Công Nghệ Hiện Đại Tổng Quan So Sánh Alipay Stack Modern Equivalent Key Difference LDC + RZone Kubernetes + Multi-cluster LDC: Business-driven sharding; K8s: Infrastructure abstraction OceanBase CockroachDB/TiDB/YugabyteDB OceanBase: 10+ years prod, custom FPGA; Newer: Cloud-native first RocketMQ Apache Kafka/Apache Pulsar RocketMQ: LSM-tree + rich msg types; Kafka: Log-centric; Pulsar: Tiered storage SOFARPC gRPC/Envoy Proxy SOFARPC: Java-centric, financial features; gRPC: Cross-platform, protobuf SOFAMesh (MOSN) Istio/Linkerd MOSN: Go-based, X-protocol; Istio: Envoy C++, standard mesh CTU Modern ML Platforms CTU: Custom fraud-specific; Modern: General-purpose MLOps PouchContainer containerd/cri-o Pouch: Alibaba-specific; containerd: CNCF standard 1. LDC Architecture vs Kubernetes Multi-Cluster Kiến Trúc So Sánh ┌─────────────────────────────────────────────────────────────────────────────┐ │ LDC (Alipay) vs Kubernetes Multi-Cluster │ ├─────────────────────────────────────────────────────────────────────────────┤ │ │ │ LDC Architecture (Business-Driven) │ │ ┌─────────────────────────────────────────────────────────────────────┐ │ │ │ │ │ │ │ RZone 1 RZone 2 RZone N │ │ │ │ ┌─────────┐ ┌─────────┐ ┌─────────┐ │ │ │ │ │Users │ │Users │ │Users │ │ │ │ │ │1-1M │ │1M-2M │ │N-M │ │ │ │ │ ├─────────┤ ├─────────┤ ├─────────┤ │ │ │ │ │Apps │ │Apps │ │Apps │ │ │ │ │ │DB │ │DB │ │DB │ │ │ │ │ │Cache │ │Cache │ │Cache │ │ │ │ │ └─────────┘ └─────────┘ └─────────┘ │ │ │ │ │ │ │ │ • Sharding: User ID-based │ │ │ │ • Self-contained units │ │ │ │ • Cross-unit = Distributed txn │ │ │ │ │ │ │ └─────────────────────────────────────────────────────────────────────┘ │ │ │ │ Kubernetes Multi-Cluster (Infrastructure-Driven) │ │ ┌─────────────────────────────────────────────────────────────────────┐ │ │ │ │ │ │ │ Cluster 1 Cluster 2 Cluster N │ │ │ │ ┌─────────┐ ┌─────────┐ ┌─────────┐ │ │ │ │ │Region: │ │Region: │ │Region: │ │ │ │ │ │us-west │ │eu-west │ │ap-south │ │ │ │ │ ├─────────┤ ├─────────┤ ├─────────┤ │ │ │ │ │K8s Pods │ │K8s Pods │ │K8s Pods │ │ │ │ │ │Services │ │Services │ │Services │ │ │ │ │ └─────────┘ └─────────┘ └─────────┘ │ │ │ │ │ │ │ │ • Sharding: Infrastructure/region-based │ │ │ │ • Shared global services │ │ │ │ • Cross-cluster = Service mesh │ │ │ │ │ │ │ └─────────────────────────────────────────────────────────────────────┘ │ │ │ └─────────────────────────────────────────────────────────────────────────────┘ Detailed Comparison Aspect LDC (Alipay) K8s Multi-Cluster Recommendation Sharding Strategy User ID / Business key Node/Region labels LDC approach cho data-intensive apps Unit Boundary App + Data + Cache Pods + Services LDC: true isolation; K8s: shared storage Cross-Unit Traffic Explicit ( costly ) Transparent via mesh LDC: intentional design; K8s: hide complexity Failover Manual/Scripted (RZone switch) Automatic (health checks) K8s wins cho automation Scaling Add RZone (complex) Add nodes (simple) K8s wins cho ops simplicity Data Consistency Strong (Paxos in unit) Eventual (cross-cluster) LDC wins cho financial data Khi Nào Dùng Cái Nào? Use LDC-style khi: ...