메뉴
CPU, memory, network bottleneck analysis — systematic performance investigation
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/fabioc-aloha/Alex_Plug_In --skill performance-profiling명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
SKILL.md 표시 중
SOC 직업 분류 기준
| name | performance-profiling |
| description | CPU, memory, network bottleneck analysis — systematic performance investigation |
| tier | standard |
| applyTo | **/*profile*,**/*performance*,**/*benchmark*,**/*bottleneck*,**/*optimize* |
Find the bottleneck before optimizing. Measure twice, optimize once.
"Premature optimization is the root of all evil" — Donald Knuth
But also:
"Measure, don't guess" — Everyone who's optimized the wrong thing
┌──────────────────────────────────────────────────────────┐
│ 1. OBSERVE │
│ User reports slowness → Reproduce → Measure baseline │
├──────────────────────────────────────────────────────────┤
│ 2. IDENTIFY │
│ Profile → Find hotspots → Determine bottleneck type │
├──────────────────────────────────────────────────────────┤
│ 3. HYPOTHESIZE │
│ Why is this slow? → Form theory → Plan fix │
├──────────────────────────────────────────────────────────┤
│ 4. OPTIMIZE │
│ Implement fix → Measure improvement → Verify │
├──────────────────────────────────────────────────────────┤
│ 5. MONITOR │
│ Add metrics → Set alerts → Prevent regression │
└──────────────────────────────────────────────────────────┘
| Type | Symptoms | Investigation |
|---|---|---|
| CPU-bound | High CPU, reasonable memory | Profile CPU, look for hot functions |
| Memory-bound | High memory, GC pauses | Heap profile, allocation tracking |
| I/O-bound | Low CPU, waiting on disk/network | Trace I/O operations, latency |
| Concurrency | Low utilization, contention | Thread dumps, lock analysis |
| Network | High latency to external services | Trace calls, measure RTT |
| Database | Slow queries, connection wait | Query plans, pool stats |
# Built-in profiler
node --prof app.js
node --prof-process isolate-*.log > processed.txt
# Chrome DevTools
node --inspect app.js
# Open chrome://inspect
# clinic.js (recommended)
npm install -g clinic
clinic doctor -- node app.js
clinic flame -- node app.js
┌─────────────────────────────────────────────────────────┐
│ main() │
│ ┌─────────────────────────┐ ┌────────────────────────┐│
│ │ processData() │ │ renderUI() ││
│ │ ┌──────────┐ ┌──────┐ │ │ ┌──────┐ ┌────────┐ ││
│ │ │ parseJSON││ │sort()│ │ │ │layout││ │ paint()│ ││
│ │ └──────────┘ └──────┘ │ │ └──────┘ └────────┘ ││
│ └─────────────────────────┘ └────────────────────────┘│
└─────────────────────────────────────────────────────────┘
Width = Time spent
# dotnet-trace
dotnet trace collect --process-id <PID> --format speedscope
# dotnet-counters (quick overview)
dotnet counters monitor --process-id <PID>
# Visual Studio Profiler
# Debug → Performance Profiler → CPU Usage
// Take heap snapshot programmatically
const v8 = require('v8');
const fs = require('fs');
const snapshotPath = `heap-${Date.now()}.heapsnapshot`;
const snapshot = v8.writeHeapSnapshot(snapshotPath);
console.log(`Heap snapshot written to ${snapshot}`);
// Memory usage check
console.log(process.memoryUsage());
// { rss, heapTotal, heapUsed, external, arrayBuffers }
| Pattern | Cause | Fix |
|---|---|---|
| Uncleared intervals | setInterval without cleanup | Store and clear in teardown |
| Event listener accumulation | Adding listeners without removing | Remove in cleanup |
| Closure capture | Large objects in closures | Null out references |
| Growing collections | Maps/Sets that never shrink | Implement eviction |
| Global state | Module-level caches | Add size limits, TTL |
# Heap dump
dotnet-dump collect -p <PID>
dotnet-dump analyze <dump-file>
# GC stats
dotnet-counters monitor --counters System.Runtime
Total Request Time: 500ms
├── DNS Lookup: 20ms
├── TCP Connection: 30ms
├── TLS Handshake: 50ms (HTTPS)
├── Time to First Byte: 350ms ← Server processing
└── Content Download: 50ms
| Issue | Symptom | Fix |
|---|---|---|
| No keep-alive | TCP handshake per request | Enable connection reuse |
| Large payloads | Slow transfer | Compress, paginate |
| No caching | Repeat downloads | Cache headers |
| Waterfall blocking | Sequential requests | Parallelize, HTTP/2 |
| DNS latency | First request slow | DNS prefetch |
-- PostgreSQL: Enable slow query logging
ALTER SYSTEM SET log_min_duration_statement = 100; -- Log queries > 100ms
-- MySQL: Enable slow query log
SET GLOBAL slow_query_log = 'ON';
SET GLOBAL long_query_time = 0.1;
-- SQL Server
-- Query Store or Extended Events
-- PostgreSQL
EXPLAIN (ANALYZE, BUFFERS, FORMAT TEXT)
SELECT * FROM orders WHERE customer_id = 123;
-- Look for:
-- - Seq Scan on large tables (need index?)
-- - High "actual rows" vs "plan rows" (stats out of date?)
-- - Nested loops with high iterations (join order?)
| Symptom | Cause | Fix |
|---|---|---|
| Waiting for connection | Pool exhausted | Increase pool size |
| Many idle connections | Pool too large | Decrease max connections |
| Connection timeout | Queries holding connections | Add query timeout |
For Alex-like extensions:
// Measure activation time
export async function activate(context: vscode.ExtensionContext) {
const start = performance.now();
// ... initialization ...
console.log(`Extension activated in ${performance.now() - start}ms`);
// Report to telemetry
telemetry.logUsage('activation', { durationMs: performance.now() - start });
}
// Wrap commands with timing
function withTiming<T>(
commandId: string,
handler: () => Promise<T>
): () => Promise<T> {
return async () => {
const start = performance.now();
try {
return await handler();
} finally {
const duration = performance.now() - start;
if (duration > 1000) {
console.warn(`Slow command: ${commandId} took ${duration}ms`);
}
}
};
}
--prof flag for detailed profiling// Node.js with Benchmark.js
import Benchmark from 'benchmark';
const suite = new Benchmark.Suite();
suite
.add('Method A', () => methodA(testData))
.add('Method B', () => methodB(testData))
.on('cycle', (event) => console.log(String(event.target)))
.on('complete', function() {
console.log('Fastest is ' + this.filter('fastest').map('name'));
})
.run({ async: true });
| Level | Latency | Example |
|---|---|---|
| L1 Cache | 1ns | CPU cache |
| L2 Cache | 4ns | CPU cache |
| RAM | 100ns | In-memory cache |
| SSD | 100μs | Local database |
| Network | 1-100ms | Remote API |
Cache strategy:
async function getCached<T>(key: string, fetchFn: () => Promise<T>, ttlMs: number): Promise<T> {
const cached = cache.get(key);
if (cached && cached.expires > Date.now()) {
return cached.value;
}
const value = await fetchFn();
cache.set(key, { value, expires: Date.now() + ttlMs });
return value;
}
// ❌ Eager - computes immediately
const allUsers = users.filter(u => u.active).map(u => u.name);
const firstTen = allUsers.slice(0, 10);
// ✅ Lazy - computes only what's needed (with generator)
function* activeUserNames(users) {
for (const user of users) {
if (user.active) yield user.name;
}
}
const iterator = activeUserNames(users);
const firstTen = Array.from({ length: 10 }, () => iterator.next().value);
// ❌ N requests
for (const id of ids) {
await fetchUser(id); // N round trips
}
// ✅ 1 request
const users = await fetchUsers(ids); // Batch endpoint
// Debounce - wait for pause in calls
function debounce(fn: Function, delay: number) {
let timeout: NodeJS.Timeout;
return (...args: any[]) => {
clearTimeout(timeout);
timeout = setTimeout(() => fn(...args), delay);
};
}
// Throttle - max once per interval
function throttle(fn: Function, interval: number) {
let lastCall = 0;
return (...args: any[]) => {
const now = Date.now();
if (now - lastCall >= interval) {
lastCall = now;
fn(...args);
}
};
}
Define acceptable limits:
| Metric | Budget | Measurement |
|---|---|---|
| Page load (LCP) | < 2.5s | Lighthouse |
| API response (P95) | < 500ms | APM |
| Memory (steady state) | < 100MB | Heap snapshot |
| Bundle size | < 200KB gzip | Build output |
| Startup time | < 100ms | Activation timing |
The fastest code is code that doesn't run. The second fastest is code that runs once.
Create and maintain ASCII visual dashboards for project tracking with parallel lane progress bars
Store and manage voice samples for TTS cloning — portable, version-controlled audio references
Clear documentation through visual excellence