Benchmark Application Performance and Optimize

This agent comprehensively benchmarks applications, identifies performance bottlenecks, and provides actionable optimization recommendations through systematic testing and analysis. It starts with system analysis: examining codebase structure to find performance-critical components, detecting the application type (web app, API, CLI tool) and tech stack, mapping dependencies and external service integrations, and establishing baseline requirements and constraints.

Test strategy design defines performance metrics (response time, throughput, resource usage, concurrency), creates test scenarios covering normal, peak, and stress conditions, chooses appropriate benchmarking tools and methodologies, and sets performance thresholds and acceptance criteria. Benchmark execution configures and runs load tests with varying user loads, executes stress tests to find breaking points, measures resource utilization (CPU, memory, disk I/O, network), tests database query performance and connection pooling, and benchmarks API endpoints and critical user journeys.

Data collection and analysis aggregates metrics across test runs, identifies bottlenecks and resource constraints, analyzes response-time distributions and outliers, and correlates performance degradation with specific components. Optimization recommendations prioritize issues by impact and implementation complexity, suggest specific code and architectural improvements, recommend infrastructure scaling strategies, and propose caching/indexing/query optimizations.

The output is a Performance Benchmark Report: an executive summary (overall rating, key bottlenecks, priority recommendations), test configuration details (environment specs, tools, scenarios executed), performance metrics (response time percentiles P50/P95/P99, throughput in RPS/TPS, resource utilization, concurrent-user capacity), bottleneck analysis with root cause and impact assessment, a phased optimization roadmap (quick wins, medium-term improvements, long-term architectural changes, infrastructure recommendations), and reusable benchmark scripts. Guidelines followed throughout: realistic testing with production-like data volumes, incremental load testing to find thresholds, environment consistency with production constraints, metrics correlated to business impact, actionable and implementable recommendations, appropriate tool selection (JMeter, wrk, ab, siege), statistically valid multi-iteration testing with confidence intervals, and resource monitoring throughout all test phases.

Use this agent when an application needs systematic performance benchmarking - finding its breaking point, measuring response time and throughput under load, and getting a prioritized optimization roadmap. It is well suited to pre-release performance validation or diagnosing a known slowdown. It is not meant for a quick manual timing check with no load testing involved - reserve it for situations where a structured, repeatable benchmark with statistical rigor is warranted.

Input: the application codebase, its type/tech stack, and target performance thresholds.

Output: a Performance Benchmark Report with metrics, bottleneck analysis, a phased optimization roadmap, and reusable benchmark scripts. Example benchmarking commands the agent uses:

# Web application load testing
wrk -t12 -c400 -d30s --script=script.lua http://example.com

# API endpoint testing
ab -n 1000 -c 50 -H "Authorization: Bearer token" http://api.example.com/endpoint

Uses standard benchmarking tools such as wrk, ab, JMeter, siege, and sysbench to load-test web applications, APIs, and databases directly.

Performance and backend engineers validating an application before release or diagnosing a known slowdown, and teams that need a repeatable, statistically valid benchmark with a prioritized optimization roadmap.