celld: Overview and Functionality

Celld is a self-hosted, distributed system designed to handle durable objects using a model that emphasizes simplicity in management and reliability. It utilizes a unique architecture that relies on a distributed coordination method without traditional consensus protocols.

Core Components and Workflow:

• Bucket as Coordinator:

  • No need for membership protocols, failure detectors, or consensus.
  • Ownership is managed through atomic writes in a bucket, ensuring that each cell's ownership is clearly defined.

• State Management:

  • Each cell functions as a SQLite database.
  • State replication is handled continuously by transmitting each cell's SQLite state to the bucket in LTX formats.

• Infrastructure Integration:

  • Celld can be deployed on dedicated virtual machines, ensuring decentralization.
  • Integrates with existing infrastructure, shipping requests and responses via a user’s infrastructure and S3 buckets.

Reliability Highlights:

• Durable Objects Model:

  • Celld incorporates this robust programming approach while decentralizing the management and operational integrity.

• Resiliency and Recovery:

  • Ensures quick recovery by relocating cell ownership through a leasing mechanism, facilitating rapid restoration without vendor dependency.

• Customization of Failure Domains:

  • Users have control over their failure domains, with no shared schedulers or placement layers that might intertwine workloads with other users.

Transparency and Accountability:

• Users have comprehensive visibility over their systems, with data and logs on their disks, enabling straightforward troubleshooting and analysis using local tools like sqlite3 and grep.

• While self-hosting doesn’t inherently increase reliability, it provides clear, manageable failure domains directly tied to the user’s infrastructure and choices.

Final Thoughts:

Celld is seen as a continuation and homage to Cloudflare’s Durable Objects model, enabling it to operate on any infrastructure that leverages technologies such as V8, SQLite, LTX, and S3.