0 views collected around this technical thread.
This paper presents the design, implementation, and experimental evaluation of OceanBase's unitization architecture for large‑scale online map services, demonstrating superior disaster‑recovery, high‑throughput OLTP/OLAP performance, and storage efficiency compared with competing distributed databases.
To meet exploding transaction volumes, rapid innovation cycles, and strict regulatory demands, large banks are replacing mainframe core systems with distributed, horizontally‑scalable architectures, and Baidu’s GaiaDB‑X database—offering strong ACID consistency, zero‑RPO disaster recovery, and automated operations—has successfully powered core banking migrations for institutions such as Bank of China and state‑owned banks.
ShopeePay replaces its traditional active‑standby design with a unit‑based multi‑active data‑center architecture—splitting user, public, and order data across independent units, employing routing tables, caches, and a dual‑active model that enables seamless failover, reduced latency, and continuous payment processing across two or more IDC sites.
The article explains multi‑active deployment modes—same‑city, cross‑city, and cross‑city multi‑active—along with routing rules (random, user‑ID, region) and unitization, describing how sharding IDs map traffic to central or unit data centers, and detailing middleware solutions for storage, messaging, SOA, and Snowflake ID generation to achieve scalable, highly available architecture.
The article explains how Ant Financial scales Double‑11 payment traffic to hundreds of thousands of TPS by employing logical data centers (LDC), unit‑based system design (RZone, GZone, CZone), database sharding, CAP theorem analysis, Paxos‑based consensus, and the OceanBase distributed database, while also detailing disaster‑recovery and traffic‑shifting mechanisms.
The article traces the evolution of backend services from monolithic applications through RPC-based scaling and sharding, highlights the limitations of database connection counts, and proposes a unitized architecture that partitions databases per service to achieve true unlimited scaling while acknowledging added complexity.
The article explains why simple sharding and database partitioning cannot achieve unlimited scaling in backend systems, analyzes the problem of excessive RPC‑to‑DB connections, and proposes a unitization approach that limits each service to a single database to enable true horizontal expansion.
This article traces the evolution of backend services from monolithic applications through RPC and sharding, explains why sharding and database partitioning cannot alone achieve unlimited scaling due to connection limits, and proposes a unitization approach that isolates database connections per service to enable true horizontal expansion.
The article explains how Ant Financial’s Logical Data Center (LDC) and unit‑based architecture, combined with sharding, CAP analysis, and OceanBase’s Paxos‑based consensus, enable the payment platform to sustain tens of millions of transactions per second during Double‑11 events while ensuring high availability and disaster recovery.
The article examines why traditional sharding and database partitioning cannot alone achieve limitless scaling, explains the problem of excessive RPC‑to‑DB connections, and proposes a unitization approach that limits each service to a single database shard to enable true horizontal expansion.
The article explains how Alipay achieves massive payment throughput during Double‑11 by using logical data centers (LDC), unit‑based system design, multi‑active disaster‑recovery, and CAP‑theorem analysis, highlighting the role of OceanBase and PAXOS in ensuring consistency and availability.
This article explains how Ant Financial’s massive Double‑11 payment traffic is handled through logical data centers (LDC), unit‑based architecture (RZone, GZone, CZone), traffic routing, disaster‑recovery strategies, and a CAP analysis that highlights the role of OceanBase’s Paxos‑based consensus in achieving high availability and eventual consistency.
The article explains how Alipay’s Logical Data Center (LDC) architecture, with its RZone, GZone, and CZone unitization, combined with OceanBase’s Paxos‑based consensus, enables massive TPS growth, traffic diversion, and disaster‑recovery while navigating the CAP theorem constraints.
The article explains how traditional monolithic, RPC, and sharding architectures eventually hit MySQL connection limits, why simple database partitioning cannot guarantee unlimited scaling, and proposes a unitization approach that restricts each service to a single database shard to overcome connection‑count bottlenecks.
The article traces the evolution of backend services from monolithic applications through RPC-based scaling and database sharding, explains why sharding alone cannot achieve unlimited scalability due to connection limits, and proposes a unitized architecture that partitions databases per service to overcome these constraints.