How to Diagnose and Optimize Business System Performance After Launch

Today we discuss the analysis, diagnosis, and optimization of business system performance issues, focusing on post‑launch problems and their remediation.

System Performance Issue Analysis Process

When a business system shows no performance problems before launch but encounters serious issues after going live, the typical causes are:

High concurrent access leading to bottlenecks

Growing database volume causing bottlenecks

Changes in key environment such as network bandwidth

First determine whether the problem appears under single‑user (non‑concurrent) conditions or only under load. Single‑user issues are usually easier to test and often stem from code or SQL, while concurrent issues require stress testing and analysis of the database and middleware.

Factors Influencing Performance

Performance factors can be grouped into hardware, software runtime environment, and application code.

Hardware Environment

Includes compute, storage, and network resources. CPU and memory capacity, as well as storage I/O, are common bottlenecks. Linux provides tools such as iostat, ps, sar, top, and vmstat for monitoring CPU, memory, JVM, and disk I/O.

Database and Middleware

Database tuning (e.g., Oracle) involves optimizing disk I/O, rollback segments, redo logs, SGA, and database objects. Middleware tuning includes configuring containers like WebLogic or Tomcat and adjusting JVM parameters.

-Xmx   # set maximum heap size
-Xms   # set initial heap size
-XX:MaxNewSize   # set maximum young generation size
-XX:NewSize      # set minimum young generation size
-XX:MaxPermSize  # set maximum permanent generation size (replaced by Metaspace)
-XX:PermSize     # set minimum permanent generation size (replaced by Metaspace)
-Xss   # set thread stack size

Recommended ratios: Xmx/Xms about 3‑4 × the live old‑generation size, Metaspace about 1.2‑1.5 ×, young generation (Xmn) about 1‑1.5 ×, and old generation 2‑3 × the live objects.

Software Code

Common code‑level performance problems include large object initialization inside loops, unreleased resources causing memory leaks, missing caching, long‑running transactions, and using sub‑optimal data structures or algorithms. Code review and static analysis are essential to detect these issues.

Performance Testing and Scaling

Pre‑release performance tests often cannot fully replicate production hardware, data volume, or concurrency, which explains why issues surface after launch. Horizontal scaling of databases is limited (e.g., Oracle RAC 2‑3×), while application clusters scale more readily, but scaling alone does not solve underlying single‑node performance defects.

Diagnosis Classification

Operating system and storage layer

Middleware layer (database, application server)

Software layer (SQL, business logic, front‑end)

APM and Monitoring

Application Performance Management (APM) monitors key business applications, improves reliability, and reduces total cost of ownership. It bridges the gap between resource‑level metrics and business‑level impact, enabling fast identification of slow services, SQL statements, or code paths.