Collaborative Text Editing with Eg-walker: Better, Faster, Smaller

Abstract: Collaborative text editing algorithms allow several users to concurrently modify a text file, and automatically merge concurrent edits into a consistent state. Existing algorithms fall in two categories: Operational Transformation (OT) algorithms are slow to merge files that have diverged substantially due to offline editing; CRDTs are slow to load and consume a lot of memory. We introduce Eg-walker, a collaboration algorithm for text that avoids these weaknesses. Compared to existing CRDTs, it consumes an order of magnitude less memory in the steady state, and loading a document from disk is orders of magnitude faster. Compared to OT, merging long-running branches is orders of magnitude faster. In the worst case, the merging performance of Eg-walker is comparable with existing CRDT algorithms. Eg-walker can be used everywhere CRDTs are used, including peer-to-peer systems without a central server. By offering performance that is competitive with centralised algorithms, our result paves the way towards the widespread adoption of peer-to-peer collaboration software.

• The paper introduces Eg-walker, a hybrid CRDT/OT algorithm that significantly reduces memory usage and merge times in collaborative editing systems.
• It benchmarks performance with real document traces, demonstrating up to 160,000× faster merges and markedly lower resource consumption than traditional methods.
• The study provides a rigorous theoretical proof of correctness, ensuring compliance with the strong list specification to support decentralized workflows.

Collaborative Text Editing with Eg-walker: Better, Faster, Smaller

The paper "Collaborative Text Editing with Eg-walker: Better, Faster, Smaller" by Joseph Gentle and Martin Kleppmann introduces Eg-walker, a novel algorithm aimed at improving the performance and efficiency of collaborative text editing systems. The research addresses existing weaknesses in the domains of Operational Transformation (OT) and Conflict-free Replicated Data Types (CRDTs) by proposing a hybrid approach that leverages strengths from both methodologies while minimizing their respective weaknesses.

Key Contributions

The paper delineates several crucial contributions:

1. Introduction of Eg-walker: Eg-walker is a hybrid CRDT/OT algorithm that achieves improved performance metrics such as CPU time for merges, steady-state memory usage, and storage size for collaborative text editing.
2. Benchmark Suite: The authors present a suite of benchmarking traces derived from real documents to evaluate collaborative text editing algorithms, thereby setting a new standard for future comparisons.
3. Performance Evaluation: The document provides comprehensive performance metrics demonstrating that Eg-walker significantly outperforms traditional OT and CRDT implementations in various scenarios.
4. Proof of Correctness: The paper includes a theoretical underpinning to prove the correctness of Eg-walker, establishing its compliance with Attiya et al.'s strong list specification.

Performance Metrics

Eg-walker showcases considerable performance improvements over existing algorithms:

• Memory Usage: In steady-state, Eg-walker utilizes 1–2 orders of magnitude less memory than the best-known CRDT implementations. This leap is primarily attributed to Eg-walker's ability to discard its internal CRDT state through the concept of critical versions, reducing the long-term memory footprint.
• Merge Time: Compared to OT, which exhibits quadratic (or worse) merge complexity, Eg-walker's merge complexity is $O(n \log n)$ in typical cases and only $O(n^2 \log n)$ in rare worst-case scenarios. This results in vastly reduced merge times, particularly in long-running branch scenarios.
• Loading Time: Eg-walker and OT both showcase loading times that are orders of magnitude faster than CRDTs. Eg-walker achieves this by caching only the final document state and relegating the event graph to disk, thus loading documents akin to loading plain text files.

Benchmark Evaluation

The benchmarks used consist of sequential, concurrent, and asynchronous traces, ensuring diverse test conditions:

• Sequential Traces: These represent the most common real-world scenarios where edits are sequential. Eg-walker excels here, showcasing performance on par with or better than OT.
• Concurrent Traces: These traces induce significant branching but for a short duration. Eg-walker demonstrates competitive performance with CRDTs while maintaining lower memory usage.
• Asynchronous Traces: These include explicit branching and merging workflows typical in version control systems. Eg-walker outperforms OT by orders of magnitude (e.g., 160,000$\times$ faster on the A2 trace) and demonstrates a slight performance edge over our reference CRDT.

Practical and Theoretical Implications

Practically, Eg-walker significantly lowers the memory overhead and enhances the responsiveness of collaborative text editors, paving the way for more resource-efficient deployment in peer-to-peer settings. Eg-walker's performance parity or superiority over traditional server-based OT algorithms opens the possibility of broader adoption of decentralized collaboration models, benefiting environments lacking reliable connectivity to central servers, such as remote fieldwork or military contexts.

Theoretically, Eg-walker reinstates the relevance of hybrid approaches in the design of distributed algorithms. By proving the correctness of Eg-walker against the strong list specification and ensuring maximal non-interleaving, the authors provide a robust mathematical foundation that can inspire further advancements in the domain of collaborative data structures.

Future Directions

Given its promising results in text editing, future research could adapt Eg-walker's framework to other collaborative data types such as rich text, graphics, or spreadsheets. Exploring enhancements in the CRDT counterpart of Eg-walker's transformation algorithm, focusing on non-interleaving guarantees, could further optimize concurrent operations. Additionally, refining the storage techniques to further minimize event graph overhead can contribute positively to the continued deployment of Eg-walker in resource-constrained environments.

Conclusion

Eg-walker represents a significant advancement in collaborative text editing algorithms. Its superior performance metrics, particularly in memory usage and CPU time for complex merges, make it a compelling alternative to existing OT and CRDT implementations. The potential for peer-to-peer collaboration it unlocks promises decentralized, resilient editing capabilities aligned with the needs of modern, distributed workflows. The thorough theoretical proofs solidify Eg-walker's standing as a key player in future collaborative systems, marking a meaningful step forward in the efficient use of distributed architectures for real-time collaboration.