liveBook · Manning

Performance optimization

As explained in Vibe Engineering

Performance optimization is the modification of software to improve responsiveness, throughput, or resource efficiency under a defined workload.

Performance optimization changes software to improve its behavior under a defined workload. Optimizing without traffic estimates or a performance target can add complexity without a production benefit. The first priority is to measure the hot path: the code doing most of the work across requests. section 7.0

A disciplined optimization loop keeps changes tied to evidence:

  • Define expected traffic and performance requirements; avoid guessing at input sizes or demand. section 7.0 section 7.5
  • Run a representative workload, measure code paths, and identify where time and work accumulate. section 7.3.1
  • Change the costly path, compare the baseline and alternative with benchmarks, and review resource and maintenance costs. section 7.1.1 section 7.4.1 section 7.5

Parallelism is not automatically faster

An account lookup already uses a lazy stream that can stop when it finds a match. Replacing it with parallelStream() introduces concurrent execution, a thread pool to monitor, and a need for stateless processing. In one benchmark, the baseline averaged 0.027 ms/op and the parallel version 0.030 ms/op; the change did not demonstrate an improvement under that workload. section 7.1.1

Measure the word-service hot path

A word service offers both daily-word retrieval and word validation. In one simulated run, validation was called 200 times and daily-word retrieval 10 times; the measurements also recorded file reads and lines scanned. Those counts and timings help prioritize work that static inspection alone can only flag as potentially expensive. section 7.2.2 section 7.3.1

Listing 7.14 The measurements implemented by LLM
{
"averageMethodTimesMs": {
  "getMetrics_endpoint": 0.638847,
  "getWordOfTheDay": 14.7273958,
  "resetMetrics_endpoint": 0.012584,
  "wordExists": 27.8176992,
  "wordExists_endpoint": 27.85579776,
  "getIndexForToday": 0.4263418,
  "getWordOfTheDay_endpoint": 14.7813709
},
"methodCallCounts": {
  "getMetrics_endpoint": 3,
  "getWordOfTheDay": 10,
  "resetMetrics_endpoint": 1,
  "wordExists": 200,
  "wordExists_endpoint": 200,
  "getIndexForToday": 10,
  "getWordOfTheDay_endpoint": 10
},
"totalFileReads": 210,
"totalBytesRead": 332209460,
"totalFileOpenTimeMs": 28.926387,
"totalStringComparisons": 31897400,
"totalLinesScanned": 32189660,
"totalRequests": 210,
"endpointCallCounts": {
  "word-of-the-day": 10,
  "word-exists": 200
},
"totalMemoryUsed": 0
}

Inspect the call counts, timings, and scanning metrics together when deciding which path matters most for this simulated workload. section 7.3.1

Caching the dictionary in a HashSet can eliminate file I/O on each validation, but the implementation matters: eager caching can waste memory for a large dictionary. Benchmark both lookup and initialization costs, and evaluate the resource trade-off rather than accepting a claimed speedup at face value. section 7.4.1

LLM assistance does not remove the need for measurement. Give it traffic assumptions and measured results before requesting changes, then review whether its proposed scope stays on the hot path. Excess instrumentation or changes across the codebase can increase analysis, review, deployment, and monitoring effort. section 7.5

How Vibe Engineering covers it

How deep it goes
Works through workload measurement, implementation alternatives, benchmarks, and trade-off review, including pitfalls in LLM-generated changes.
Examples
Compares sequential and parallel account lookup and instruments a word service before caching its dictionary.
What you can do after
Evaluate optimization proposals against workload measurements, benchmark results, and operational trade-offs.
How it connects
Connects performance optimization to profiling, benchmarking, parallelism, caching, and service-level agreements.

Also explained in

Compare how 10 books explain performance optimization →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Vibe Engineering. The section links point to the parts of the book it draws on.