AGENT SKILL

performance-tuning

Use when investigating slow ABAP programs, expensive SQL statements, or reviewing performance risk on large tables (VBAK, BSEG, ACDOCA, etc.). Provides a standardized workflow using TraceExecution, ListSQLTraces, and GetCallGraph to diagnose and document performance issues before they reach production.

License 许可
AGPL-3.0
最近更新
2026-11-14
Stars
★ 25
Forks
9

01 前言 Front Matter

name
performance-tuning
owner
dba
scope
co-abap
status
active
version
1.0.0
metadata
{ "type": "core", "triggers": [ "performance-tuning", "TraceExecution", "ListSQLTraces", "GetCallGraph", "slow program" ] }
description
Use when investigating slow ABAP programs, expensive SQL statements, or reviewing performance risk on large tables (VBAK, BSEG, ACDOCA, etc.). Provides a standardized workflow using TraceExecution, ListSQLTraces, and GetCallGraph to diagnose and document performance issues before they reach production.
last_reviewed
1786752000000
prerequisites
vsp MCP server

03 SKILL.md 正文

Performance Tuning Workflow

Standardizes performance analysis so findings are reproducible and comparable across sessions, instead of ad-hoc RunQuery sampling. Owned by the DBA agent (agents/dba.md); the Architect and QA Engineer may trigger it during technical design or QA review.

Origin: raised by the DBA agent in the 2026-07-05 project review meeting — large-table access paths (VBAK, BSEG) need a standard procedure, not one-off spot checks.

When to Use

  • A custom program or CDS view queries a large/high-traffic table (VBAK, VBAP, BSEG, ACDOCA, MSEG, EKPO)
  • QA or the user reports a slow transaction, report, or OData service
  • Technical Design (Stage 2) for any object expected to run against high-volume tables
  • Before releasing a transport that touches performance-sensitive objects

Workflow

1. Baseline trace
   TraceExecution(object) → capture runtime, DB time %, top statements

2. SQL statement analysis
   ListSQLTraces() → GetSQLTraceState() → identify statements with high execution count
   or full table scans (missing index usage)

3. Call graph / dependency check
   GetCallGraph(object) or AnalyzeCallGraph(object) → confirm no redundant nested SELECTs
   inside loops (SELECT-in-LOOP anti-pattern)

4. Index verification (hand off to DBA table-design duties)
   GetTable(table) → check existing secondary indexes cover the WHERE/JOIN predicates
   used by the traced statements

5. Report
   Document findings using the Performance Report format below

Performance Report Format

### Performance Analysis — <object name>

**Trigger**: <why this was analyzed — slow report / pre-release check / QA escalation>
**Trace summary**: <total runtime, DB time %, top 3 statements by execution count>

#### Findings
| # | Statement / Pattern | Table(s) | Issue | Severity |
|---|---------------------|----------|-------|----------|
| 1 | ... | VBAK | Full table scan (missing index on ERDAT) | High |

#### Recommendations
- <index addition / query rewrite / SELECT-in-LOOP fix, one per finding>

**Verdict**: <Pass / Needs optimization before transport release>

Anti-Patterns to Flag

  • SELECT inside a LOOP over another internal table (N+1 pattern) — rewrite as a single SELECT ... FOR ALL ENTRIES or JOIN.
  • Missing WHERE clause filtering on large tables before RunQuery/SELECT.
  • Full table scans on VBAK, VBAP, BSEG, ACDOCA without a covering secondary index.
  • Nested CDS view associations that re-resolve the same large table per row.

Severity → Disposition

Severity Disposition
High (full scan / N+1 on a large table) Block transport release until fixed
Medium (missing index, acceptable volume today) PM disposition — fix now or track as follow-up task
Low (style-only inefficiency, negligible volume) Note in QA report, no gate

Related

Context

This skill standardizes performance analysis for ABAP programs and SQL queries so findings are reproducible and comparable across sessions. It replaces ad-hoc RunQuery sampling with a structured workflow using TraceExecution, ListSQLTraces, and GetCallGraph to diagnose bottlenecks and document recommendations before they reach production. It is owned by the DBA agent but may be triggered by Architect or QA Engineer during design or review.

Execution Steps

  1. Baseline trace — Use TraceExecution on the target object to capture runtime, DB time percentage, and top statements.
  2. SQL statement analysis — Use ListSQLTraces and GetSQLTraceState to identify high-execution-count statements or full table scans.
  3. Call graph check — Use GetCallGraph or AnalyzeCallGraph to detect redundant nested SELECTs inside loops (N+1 anti-pattern).
  4. Index verification — Use GetTable to check whether existing secondary indexes cover the WHERE/JOIN predicates in traced statements.
  5. Report — Document findings using the standardized Performance Report format with findings table and recommendations.

Output Format

A structured Performance Analysis Report containing:

  • Trigger reason (slow report, pre-release check, QA escalation)
  • Trace summary with total runtime, DB time percentage, and top statements
  • Findings table with statement/pattern, tables affected, issue description, and severity
  • Recommendations for each finding (index addition, query rewrite, or code fix)
  • Verdict (Pass or Needs optimization before transport release)

Related Skills

  • abap-dev — Core ABAP development workflows including the performance-analyzer sub-section
  • post-write-chain — Run after performance fixes pass to validate no regressions
  • dump-monitor — Use together when a dump indicates a performance-related timeout

04 同仓库更多 Skills

5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25