ClickHouse claims its WAL-RUS rewrite is 2x faster for PostgreSQL backups

ClickHouse has rewritten the Go-based WAL-G in Rust, creating WAL-RUS. The new tool promises major performance gains for PostgreSQL backup and restore operations, which we analyze here.

The Answer Up Front

For engineering teams managing large-scale PostgreSQL deployments where backup and restore times are a primary operational bottleneck, WAL-RUS is a compelling new option that warrants immediate evaluation. For most other users, particularly those with smaller databases or who prioritize ecosystem maturity and stability, the battle-tested WAL-G remains the safer default choice. The bottom line is that WAL-RUS is a high-performance, specialized tool for solving a specific scale problem. Its adoption depends entirely on whether backup performance is a first-order pain point for your team.

Methodology

This v0 review analyzes the initial announcement and performance claims for WAL-RUS, published by ClickHouse on June 28, 2026. The analysis is based exclusively on the technical details and benchmark results presented in the source blog post, "WAL-RUS: a Rust Rewrite of WAL-G for PostgreSQL Backups." We have not conducted independent performance benchmarks or tested the tool in a production environment. This review covers the architectural rationale for the rewrite from Go to Rust, the specific performance claims made for backup and restore operations, and the tool's intended compatibility with existing WAL-G setups. It does not cover long-term stability, resource consumption (CPU/memory) under varied loads, or behavior in edge-case scenarios like unstable network conditions. All performance metrics cited here are claims from ClickHouse, pending independent verification.

What It Does

A drop-in replacement for WAL-G

WAL-RUS is designed to be a direct, command-compatible replacement for WAL-G, a widely used open-source tool for PostgreSQL continuous archiving and backups. The explicit goal is to allow existing WAL-G users to switch with minimal configuration changes. It interacts with the same storage backends (like S3, GCS, and Azure Blob Storage) and maintains the same on-disk backup format. This ensures that a backup taken with WAL-G could theoretically be restored with WAL-RUS, and vice-versa, lowering the barrier to adoption.

Performance-focused architecture

The primary motivation for rewriting WAL-G in Rust was performance. The ClickHouse team identifies several architectural factors contributing to the speedup. They claim Rust's lack of a garbage collector reduces performance stalls and overhead compared to Go. The implementation also uses io_uring for asynchronous I/O on Linux, which can be more efficient than traditional blocking I/O models. The post also points to fine-grained control over memory layout and concurrency as key advantages of the Rust implementation, allowing them to optimize data pipelines for encryption, compression, and network transfer more effectively.

Core backup and restore commands

Functionally, WAL-RUS provides the same core utilities as its predecessor. The key commands covered in the announcement are:

• backup-push: Creates and uploads a full database backup to remote storage.
• backup-fetch: Downloads and restores a full backup.
• wal-push: Archives a single Write-Ahead Log (WAL) segment.
• wal-fetch: Retrieves a single WAL segment for point-in-time recovery.

The performance claims center on the backup-push and backup-fetch operations, which handle the bulk data transfer.

What's Interesting / What's Not

The most interesting aspect is the stark performance claim. ClickHouse reports that in their benchmark environment (using a 1TB dataset on an AWS EC2 instance), WAL-RUS completed a backup-push operation up to 2x faster than WAL-G. The backup-fetch operation was reportedly up to 1.8x faster. These are significant improvements for teams where a multi-hour restore time is a major liability. The provided benchmark graphs are specific: the big gains are in bulk data operations. For incremental wal-push and wal-fetch operations, the performance difference is shown to be minimal, which makes sense as these are smaller, latency-sensitive tasks rather than throughput-bound ones. This detail is crucial; WAL-RUS won't magically speed up all aspects of your backup system.

What's not particularly novel is the tool's feature set. By design, it's an optimization play, not an innovation in backup strategy. It doesn't introduce new backup methods or change the fundamental WAL-G workflow. This is a deliberate, pragmatic choice to encourage adoption. However, it means the decision to switch is almost entirely a performance calculation. The project also highlights a broader industry trend: rewriting critical infrastructure tooling in Rust for performance and memory safety, trading the developer convenience of Go for lower-level system control.

Pricing

WAL-RUS is an open-source tool released under the Apache 2.0 License. It is free to use. (Pricing checked June 28, 2026).

Verdict

WAL-RUS appears to be a successful execution of a targeted performance optimization. For organizations running PostgreSQL at a scale where backup and restore operations are measured in hours and directly threaten Recovery Time Objectives (RTOs), it is a compelling alternative to WAL-G. The promise of halving the time for a full restore is a powerful incentive. However, for the majority of PostgreSQL users, WAL-G's maturity, larger community, and proven track record provide a level of stability that a brand-new rewrite cannot yet claim. The choice is clear: if you have a demonstrated, measured performance problem with your current backup tooling, benchmark WAL-RUS. If not, stick with the incumbent.

What We'd Test Next

A v2 review would require independent benchmarks. First, we would replicate the ClickHouse benchmark on identical hardware to verify the 2x claim. Next, we would test across different cloud storage providers and network conditions to see how performance varies. We would also measure CPU and memory consumption during backup-push and backup-fetch operations, as wall-clock time is only one component of cost and performance. Finally, a long-term correctness and stability test would be essential, involving months of accumulated WAL segments to ensure point-in-time recovery works flawlessly under real-world conditions.

The investor read

WAL-RUS is not a standalone company, but a strategic open-source project by ClickHouse. Its significance lies in what it signals about the infrastructure software market. First, it affirms the trend of Rust supplanting Go for performance-critical tooling where squeezing out maximum efficiency is paramount. This creates opportunities for new entrants to disrupt incumbents by competing on performance rather than features. Second, for ClickHouse, this is a savvy ecosystem play. By contributing a high-value tool to the PostgreSQL community, they build brand equity and goodwill, attracting engineering talent and mindshare among developers who might also use their core database product. This isn't a direct investment opportunity, but a barometer for where performance-engineering-driven value is being created in the open-source data infrastructure space.