Performance Engineering

Perf Hub

Exploring the art and science of building fast, scalable, and reliable systems.

// Intro

What is Performance Engineering?

Performance Engineering is a discipline within systems engineering that focuses on ensuring a system will meet or exceed its performance requirements. It's not just about testing; it's a proactive process integrated throughout the entire software development lifecycle.

This hub is a collection of my notes, case studies, and thoughts on key concepts, tools, and methodologies in the world of performance.

// 01

Core Concepts

01

Latency & Response Time

The time taken for a single transaction to be processed.

02

Throughput

The number of transactions a system can handle per unit of time (e.g., requests/second).

03

Scalability

The ability of a system to handle a growing amount of work by adding resources.

04

Capacity Planning

The process of determining the production capacity needed by an organization to meet changing demands for its products.

05

Workload Modelling

Creating a representative model of user and system load to be used in performance tests.

// 02

Performance Workload Modelling

Illustrative workload models showing how business scenarios turn into load profiles and pass/fail criteria. Sample figures, not client data.

E-commerce Peak Sale Event

Load50,000 concurrent
Throughput1,200 transactions/sec
Avg response250ms
Peak800ms (95th percentile)
Response Time by Operation (ms)
150Login200Search180Add to Cart450Checkout
  • User login and session management
  • Product search and filtering
  • Shopping cart operations
  • Secure checkout and payment processing

High-Traffic API Gateway

LoadN/A (Service-to-Service)
Throughput10,000 requests/sec
Avg response50ms
Peak150ms (99th percentile)
Latency Distribution (ms)
45p5080p90110p95150p99250p99.9
  • Authentication and authorization (JWT/OAuth)
  • Request routing and payload transformation
  • Rate limiting and caching strategies
// 03

Essential Tooling

Load Generation

JMeterGatlingk6Locust

Monitoring & Observability

PrometheusGrafanaDatadogNew RelicOpenTelemetry

Profiling

JProfilerVisualVMpy-spypprof
// 04

Case Studies & Analyses

CASE STUDY

Optimizing E-commerce Checkout API

A deep dive into identifying and resolving a critical bottleneck in a high-traffic checkout service. The project involved profiling, query optimization, and implementing an asynchronous processing pattern.

  • Reduced p95 latency from 2500ms to 400ms.
  • Increased transaction throughput by 300% during peak load.
  • Eliminated database lock contention issues.
// 05

Calculators

Quick sizing tools I reach for in test planning. They run in your browser and send nothing anywhere.

Little's Law

Users = throughput × (response time + think time)

Concurrent users0

Pacing

Pacing = users × 3600 ÷ iterations per hour

Pacing interval0 s
Wait after each iteration0 s
Throughput0 /s

Peak load from daily volume

Peak TPS = daily volume × busiest-hour share ÷ 3600

Average TPS0
Peak-hour TPS0
Test target with factor0

Apdex score

(Satisfied + Tolerating ÷ 2) ÷ total samples

Apdex0
Rating-

Availability and downtime

What an availability target allows

Per year-
Per month-
Per week-
Per day-

Percentiles

Paste response times (any separator). Nearest-rank method.

Samples0
Average-
Min / Max-
P50-
P90-
P95-
P99-
// 06

Compare utilities

Paste two versions and see what changed. Text and Java use a line diff, JSON and YAML compare the structure, so key order and formatting do not matter. Everything stays in your browser.

Result appears here
// 07

Dev utilities

Small helpers for everyday debugging. Nothing is sent anywhere, so it is safe to paste tokens and payloads.

JSON formatter

Validate, pretty-print or minify

Base64

Encode or decode, UTF-8 safe

URL encode / decode

Useful for query strings and correlation values

JWT decoder

Reads header and payload. The signature is not verified.

Regex tester

JavaScript regex with live matches

Epoch time

Seconds or milliseconds to date, and back