Hello everyone, today let's discuss a topic that many people are paying close attention to: how can the decentralized storage field break through?

Frankly speaking, the core contradiction in this track has never been about "whether storage is possible," but rather about balancing three dimensions—cost, recoverability, and anti-malicious behavior. Achieving all three simultaneously? Easier said than done.

Copy data completely? Excellent stability, but the cost is so high that it makes you question life. Traditional one-dimensional erasure coding schemes (like Reed-Solomon) have brought prices down, which is true, but problems also follow: if nodes frequently go offline in your network environment—which is common in real-world applications—restoring a data fragment requires re-downloading and recalculating the entire file. Bandwidth? Explodes. Time? Also explodes. User experience? Even more so.

Walrus's approach is different. They don't intend to choose between cheap and reliable as an either/or option. Instead, they decided to redesign the underlying coding logic, enabling the system to inherently possess self-healing capabilities. The core of this solution is called RedStuff.

According to official documentation, RedStuff is a custom erasure coding scheme developed by Walrus, based on the mathematical foundation of "efficient and computable Reed-Solomon codes," while achieving three key properties: systematization, determinism, and high efficiency.

What does systematization mean? Simply put, the data received by some storage nodes isn't a scrambled or encrypted fragment, but a direct slice of the original data. The practical benefit is clear: common operations like random reads and range queries no longer require full decoding each time, significantly improving performance.

Determinism guarantees security from another perspective. The encoding process for the same data has no "human freedom"—regardless of who encodes or when it is encoded, the result must be exactly the same. This effectively prevents nodes from maliciously tampering with data through "creative operations" during encoding.

And those two eye-catching numbers further illustrate the point: Walrus's system only needs about one-third of the redundancy to reach the security threshold, and during data repair, it can reduce the computational load by approximately three times compared to traditional schemes. In other words, under the same security guarantees, costs are lower, and speed is faster.

The brilliance of this approach lies in its not optimizing along a single dimension but fundamentally improving the entire system's efficiency boundary by changing the organization of the underlying coding. In the path of decentralized storage, such architectural-level innovation is often more valuable than mere engineering optimizations.