⚡ The Performance Bible
The Architect's Handbook: From Code to Physics
"The language is just a tool. You don't blame the hammer if you can't build a skyscraper alone."
📚 30 Topics
📖 800+ Pages Equivalent
💻 200+ Code Examples
👤 EzEldeen A. Y. Mushtaha
Software Architect & Performance Engineer
📅 Originally compiled: 2026 • All content based on real-world production experience
Topic 02
Route Caching
Eliminate route parsing overhead on every request.
- Without cache: 200 routes = 30ms regex parsing per request
- With cache: routes precompiled as PHP array → 2ms
- Command: php artisan route:cache
- Part of php artisan optimize
📖 Read Topic 02 →
Topic 03
Config Caching (env trap)
The #1 production killer after config:cache.
- Never use env() outside config/*.php files
- After config:cache, env() returns null
- Always use config() in your code
- Saves 15-40ms per request + eliminates I/O
📖 Read Topic 03 →
Topic 04
Eloquent N+1 Killer
The #1 performance disaster in Laravel.
- 101 queries when 2 would suffice
- Solution: with() for eager loading
- preventLazyLoading() catches N+1 in development
- Hidden N+1 in views (the real trap)
📖 Read Topic 04 →
Topic 05
Hydration Hell
Eloquent vs Query Builder memory comparison.
- Eloquent: 500MB for 100k users
- Query Builder: 50MB for 100k users
- Read-only? Use DB::table() or raw PDO
- 5-10x memory difference
📖 Read Topic 05 →
Topic 06
Queue Serialization Trap
Why sending full models to queues kills Redis memory.
- Full model serialization: 5KB per job
- ID only: 8 bytes per job → 99.9% reduction
- Always dispatch $user->id, not $user
- Fetch fresh data inside the job
📖 Read Topic 06 →
Topic 07
Service Container & Reflection
Why auto-wiring is expensive and how to fix it.
- Reflection = code reading code at runtime (10-100x slower)
- Singletons for stateless services
- Explicit binding to skip reflection
- Never call make() inside loops
📖 Read Topic 07 →
Topic 08
Middleware Selectivity
Every request runs 10 middleware classes. It shouldn't.
- Health check doesn't need session, auth, or CSRF
- Create specialized groups: public, auth-api, web-session
- Order matters: put fast-failing middleware first
- Save 5-10ms per request
📖 Read Topic 08 →
Topic 09
Database Indexing & EXPLAIN
The #1 database performance killer.
- Missing index = full table scan = 5 seconds for 1M rows
- With index = 0.001 seconds
- Run EXPLAIN before any query
- Composite index order matters (left-most prefix)
📖 Read Topic 09 →
Topic 10
Read/Write Splitting
Master for writes, replicas for reads.
- 90% of traffic is reads. Replicas handle them.
- Laravel config: no code changes needed
- Double your capacity overnight
- Beware of replication lag (eventual consistency)
📖 Read Topic 10 →
Topic 11
Database Partitioning
Vertical & Horizontal partitioning.
- Vertical: split large columns (BLOB, TEXT) to separate tables
- Horizontal: split by date or user_id (sharding)
- MySQL native partitioning (by date, DROP PARTITION in seconds)
- Application-level sharding for multi-server scale
📖 Read Topic 11 →
Topic 12
Statelessness Doctrine
Your server must be killable at any moment.
- Sessions → Redis (not files)
- Files → S3 (not local disk)
- Logs → CloudWatch / ELK
- Config → Environment Variables
📖 Read Topic 12 →
Topic 13
Event-Driven Architecture
Stop making users wait. Process in the background.
- User waits: 500ms → 20ms (96% faster)
- Dispatch events, return 202 Accepted
- Queues + ShouldQueue for background processing
- Scale workers independently
📖 Read Topic 13 →
Topic 14
CDN & Edge Caching
Keep Laravel away from static files.
- Nginx serves static files (1ms vs Laravel 50ms)
- CDN caches globally (Tokyo: 150ms → 5ms)
- Browser caching with Cache-Control headers
- Asset versioning for cache busting
📖 Read Topic 14 →
Topic 15
The 10 Disasters
The most common production failures.
- DB::enableQueryLog() in production → OOM
- env() after config:cache → returns null
- User::all() for large datasets → memory explosion
- Full model in queue → payload bloat
- Forgotten optimize → 30ms waste per request
- Lazy loading N+1 in views
- Log::info() inside loops → I/O death
- File session driver → concurrency killer
- Missing indexes → full table scan
- pm.max_requests = 0 → memory leaks
📖 Read Topic 15 →
Topic 16
PHP-FPM Deep Dive
The real bottleneck nobody talks about.
- pm.max_children calculation: (RAM - others) / 40MB
- pm.max_requests = 500 (prevents memory leaks)
- dynamic vs ondemand vs static
- request_terminate_timeout = 60s
📖 Read Topic 16 →
Topic 17
OpCache & PHP JIT
PHP's built-in performance superpowers.
- OpCache: compile once, cache forever (3-5x faster)
- JIT: compile hot paths to machine code (PHP 8+)
- OpCache configuration for production
- JIT for CPU-intensive apps only
📖 Read Topic 17 →
Topic 18
Composer Autoloader
Stop PHP from searching for files.
- --classmap-authoritative eliminates filesystem search
- APCu autoloader caching: classmap stays in memory
- Full production command: composer install --no-dev --optimize-autoloader --classmap-authoritative
- 15-20ms → 2-4ms per request
📖 Read Topic 18 →
Topic 19
Database Proxy (ProxySQL)
Connection pooling for MySQL.
- Eliminate TCP handshake per request (1.5ms → 0.1ms)
- Reuse connections across requests
- Query caching in ProxySQL
- Read/write splitting at proxy level
📖 Read Topic 19 →
Topic 20
APCu vs Redis
Local shared memory vs network-based cache.
- APCu: 0.05µs, Redis: 10µs (200x faster)
- APCu for config, feature flags, local counters
- Redis for sessions, queues, shared cache
- Hybrid approach: APCu + Redis layered cache
📖 Read Topic 20 →
Topic 21
Laravel Octane
The nuclear option. Keep Laravel alive in memory.
- Swoole vs RoadRunner
- Eliminate bootstrap overhead → 10x faster
- Memory leak danger in persistent processes
- Octane-safe code patterns
📖 Read Topic 21 →
Topic 22
SplFixedArray
PHP arrays are hash maps. SplFixedArray is a true C array.
- PHP array: 500MB for 1M integers
- SplFixedArray: 100MB for 1M integers (80% less)
- Fixed size, integer keys only
- 20-40% faster than normal arrays
📖 Read Topic 22 →
Topic 23
Lazy Resolution for Services
Don't pay for services you don't use.
- Eager: all services instantiated on every request
- Lazy: instantiate only when used
- Bind with class name strings for lazy resolution
- 70-80% reduction in service overhead
📖 Read Topic 23 →
Topic 24
Observers & withoutEvents
The silent performance killers.
- Without events: bulk update = 5ms
- With observers: bulk update = 5,000ms (1000x slower)
- Use withoutEvents() for bulk operations
- Move observer logic to queued jobs
📖 Read Topic 24 →
Topic 25
Logging I/O Trap
Every Log::info() opens a file.
- 10,000 logs = 10,000 I/O operations (1-2 seconds)
- Batch logs in arrays: 2 I/O operations (0.002 seconds)
- Set LOG_LEVEL=error in production
- Use Redis logging for high volume
📖 Read Topic 25 →
Topic 26
Blade View Cache
Every Blade template is compiled to PHP.
- Without cache: 5-20ms per view per request
- With view:cache: 1-2ms per view
- Command: php artisan view:cache
- Part of php artisan optimize
📖 Read Topic 26 →
Topic 27
CAP Theorem
Consistency vs Availability vs Partition Tolerance.
- You can only have two of three
- CP (banking): consistency over availability
- AP (social media): availability over consistency
- CA is impossible in distributed systems
📖 Read Topic 27 →
Topic 28
Distributed Locking
Cache::lock is not enough for multi-server environments.
- Redlock: 5 Redis instances for safety
- ZooKeeper: strong consistency, fencing tokens
- etcd: modern alternative (Kubernetes)
- MySQL GET_LOCK: simple but slow
📖 Read Topic 28 →
Topic 29
Observability
Logging, Metrics, Tracing — the three pillars.
- Logs: what happened (ELK, Loki)
- Metrics: how often (Prometheus, Datadog)
- Tracing: where time went (Jaeger, Tempo)
- Correlation IDs connect everything
📖 Read Topic 29 →
Topic 30
Zero-Trust Architecture
Trust nothing. Verify everything.
- Death of the perimeter (no inside/outside)
- Authenticate every request, every service
- Least privilege, JIT access, micro-segmentation
- Assume breach, monitor continuously
📖 Read Topic 30 →