Cluster Laravel Package

Product Decisions This Supports

• Roadmap: avoid or defer a language rewrite. If the product needs more concurrency for PHP-based real-time or network services, this package supports evaluating “scale the PHP/AMPHP service” before committing to a Go/Node migration.
• Architecture: scale-up vs scale-out. Informs whether to use multiple worker processes per host to better use CPU cores before adding more pods/hosts, especially for long-lived connections and event-driven network services.
• Build vs adopt. Supports adopting an existing MIT worker-cluster layer instead of building custom master/worker process orchestration for AMPHP services. It is infrastructure plumbing, not product functionality.
• Feature use cases. Relevant to WebSocket/SSE gateways, chat/presence, notification fan-out, streaming APIs, internal TCP services, and other AMPHP-based network daemons. Less relevant to standard Laravel CRUD, queue workers, or request/response apps.
• Investment and risk decisions. Helps decide whether a time-boxed spike is worth funding, what operational ownership is required, and whether the dependency is acceptable for a critical path. Package signals are niche: 61 stars, last release Dec 2024, low opportunity score — treat as tactical, not a platform bet.
• Laravel-specific caveat. This is not a Laravel-native package and is not a drop-in dependency for typical Laravel apps; it requires an AMPHP-style async runtime and process model.

When to Consider This Package

Consider it when:

• The team already uses or is committed to AMPHP (amphp/http-server, WebSocket/socket libraries, etc.).
• A benchmark shows a single-process async PHP service is hitting a concurrency/CPU-utilization ceiling and multiple worker processes are a plausible fix.
• The service can keep workers stateless, with shared state in Redis, a database, or a queue.
• Engineering can own process supervision, graceful restarts, health checks, logging, metrics, and crash recovery.
• The team accepts a small-community OSS dependency with MIT licensing, no commercial SLA, and the need to pin/review versions.
• There is a clear fallback: stay with horizontal scaling, use Laravel Octane/RoadRunner/FrankenPHP/Swoole, or move the bottleneck to dedicated workers.

Look elsewhere when:

• The app is a standard Laravel CRUD/API app or uses queue workers for background jobs.
• The stack is not AMPHP-based, or the team lacks async PHP/process-model experience.
• You need enterprise support, broad community validation, or a low-risk critical-path dependency.
• The problem is CPU-bound computation rather than network concurrency; use job workers or native services instead.
• You cannot justify the operational overhead of multi-process debugging and supervision.
• Cross-worker coordination is required; use a queue/pub-sub system rather than relying on cluster workers for shared state.

How to Pitch It (Stakeholders)

Executive summary: amphp/cluster is a niche infrastructure package that lets AMPHP-based PHP network services use multiple CPU cores through worker processes. It does not add user-facing features; it may extend the life of a PHP async service and reduce pressure to rewrite in another language. Given its small community and limited adoption signals, it should be positioned as a time-boxed technical spike for a proven bottleneck — not a platform-wide standard.

Engineering summary: Use it only if we are already on AMPHP and need worker-process orchestration rather than building it ourselves. Keep workers stateless, externalize shared state, and budget for supervision, observability, graceful shutdown, and upgrade risk. Success criteria should be explicit: measured throughput/latency improvement on a realistic workload, a runbook for operations, and a clear exit path if the dependency or complexity does not pay off.

Recommendation: Conditional pilot, not broad adoption. Fund a short spike against one concrete service bottleneck; if it does not beat simpler scale-out or Octane/RoadRunner/FrankenPHP alternatives on cost, risk, and operability, do not adopt.