research(storage): prototype checkpoint ring buffer backend

Problem

The current memory storage implementation uses standard synchronization primitives to protect workflow state.

While this provides correctness, high-frequency checkpoint operations may benefit from exploring alternative memory layouts optimized for sequential state access patterns.

---

Proposed Solution

Prototype a ring buffer based checkpoint storage backend.

The implementation should:

• Create an experimental ring buffer storage component.
• Maintain checkpoint ordering using head and tail positions.
• Explore safe concurrent access patterns.
• Compare performance against the existing MemoryStore.
• Add benchmarks for throughput and memory behavior.

Goals:

• Evaluate alternative storage layouts.
• Understand performance trade-offs.
• Keep the current storage implementation unchanged.

---

Alternatives

Alternative approaches considered:

• Continue using map-based storage with mutex protection.

  • Simple and reliable but may limit experimentation with high-frequency workloads.

• Introduce external databases immediately.

  • Adds operational complexity before storage requirements are fully defined.

---

Use Case

A ring buffer experiment can help evaluate future optimizations for:

• High-frequency checkpoints
• Local state buffering
• Runtime recovery pipelines
• Distributed synchronization preparation

---

Additional Notes

Requirements:

• This is an experimental backend.
• Existing storage interfaces must remain compatible.
• Production storage behavior should not change.
• Include benchmark comparisons.

Related areas:

• internal/storage/
• internal/control/checkpoint.go

Start by reading the existing storage implementation under internal/storage/ and the checkpoint flow in internal/control/checkpoint.go. Identify the current MemoryStore interface and its compatibility requirements before designing the experimental backend. Done means a ring buffer prototype preserves checkpoint ordering, includes safe concurrent access, and has benchmark comparisons for throughput and memory behavior without changing production storage.