When traffic spikes, a multi‑table JOIN can hold database connections far longer than separate single‑table queries, quickly exhausting the connection pool and slowing the entire system, as demonstrated by concrete timing examples and a practical workaround.
This article provides a step‑by‑step tutorial for configuring Apache Superset to connect to a wide range of databases, covering driver installation, Docker image customization, exact SQLAlchemy connection strings for each engine, UI‑based setup, advanced options such as SSL, external password stores, and the experimental meta‑database feature.
The article explains that although IO multiplexing can improve performance, database access in Java typically relies on JDBC and connection pools, which are built on blocking I/O, making it difficult to combine DB connections with IO multiplexing without major architectural changes.
This article shares practical lessons from optimizing the performance of legacy web applications, covering how a hidden DB‑connection leak caused severe latency, why average latency is misleading, and the essential tools, testing, and team practices needed to keep services fast and reliable.
After a sudden traffic surge exposed severe latency in a 15‑year‑old multi‑service web system, the team identified a DB connection leak in a health‑check probe, and through load testing, monitoring, logging, and dedicated SRE effort, they derived key operational lessons on performance optimization and maintenance of legacy applications.
This article shares hard‑won lessons from optimizing fifteen high‑load web applications, highlighting how a tiny DB‑connection leak in a pod probe caused severe slowdown and how fixing it, along with proper load testing, monitoring, and investment in tools and people, can dramatically improve system performance.
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.
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 why traditional sharding and database partitioning cannot guarantee limitless scalability, examines the connection‑limit problems of RPC‑based services, and introduces unitization as a practical approach to achieve true unlimited scaling while managing database connections efficiently.
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.
