Performance Optimization
Overview
Measure first, optimize second. Never optimize without data. The bottleneck is almost never where you think it is.
When to Use
- Code is measurably slow or resource-heavy
- Profiling has revealed specific bottlenecks
- Designing a system with known performance requirements
- Reviewing code for performance before production
When NOT to Use
- You "feel like" code might be slow but haven't measured
- Premature optimization during initial implementation
- Micro-optimizations that don't move the needle
The Process
1. Measure baseline Before touching anything, establish a benchmark.
- Record current performance numbers (time, memory, CPU)
- Document the test conditions (data size, concurrency, hardware)
- Save results as your baseline
2. Profile to find the bottleneck Use profiling tools appropriate to your stack:
- Python:
cProfile,py-spy,memory_profiler - Node.js:
--prof,clinic.js, Chrome DevTools - Go:
pprof - Generic: timing instrumentation, APM tools
The bottleneck is the one place where optimization actually matters.
3. Form hypothesis State explicitly: "I believe X is slow because Y." Don't optimize something you can't explain.
4. Apply targeted fix Change ONE thing at a time. Common high-impact areas:
- Database: N+1 queries, missing indexes, over-fetching
- Network: unnecessary round trips, large payloads, no caching
- Memory: leaks, excessive allocation, large objects in hot paths
- Algorithms: O(n²) where O(n log n) is possible
- I/O: synchronous blocking, missing batching
5. Measure again Compare to baseline. Did it improve? If not, revert and try something else.
6. Document the change Record what you changed, why, and the before/after numbers.
Common High-Impact Wins
| Area | Look For |
|---|---|
| Database | N+1 queries, full table scans, missing indexes |
| Caching | Repeated expensive computations with same inputs |
| Network | Chatty APIs, large payloads, synchronous chains |
| Algorithms | Nested loops over large collections |
| Memory | Objects created in tight loops, large in-memory datasets |
Red Flags
| Thought | Reality |
|---|---|
| "This looks slow" | Measure it. Looks are deceiving. |
| "I'll optimize as I go" | Premature optimization obscures intent. Measure first. |
| "I fixed the bottleneck" | Did you measure? Fix without measurement isn't a fix. |
| "This is the obvious bottleneck" | Profile anyway. You're probably wrong. |