I spent nearly two weeks going through this decentralized storage solution based on the Sui chain from start to finish. To be honest, it has many highlights but also some real-world pitfalls.

When it launched on the mainnet late last year, it caused quite a stir. Its core selling point was using Red Stuff erasure coding technology to reduce storage costs to a fraction of competing products. This technology indeed attracted a wave of developers to try it out. But only when you get hands-on can you see the intricacies of the system architecture, and it also exposes some stage-specific challenges.

From a technical perspective, the core is a combination of fast linear fountain codes with Byzantine fault tolerance mechanisms. Put simply, it involves breaking files into shards, generating redundant data through algorithms, and dispersing them across various storage nodes. The cleverness of this approach is that— in theory— only 4 to 5 times redundancy is needed to ensure data integrity and recoverability. In comparison, some leading projects use full network replication (equivalent to hundreds of times redundancy), and Filecoin often employs a 25x replication factor, which makes the cost difference immediately apparent.

To put it in numbers: a 100MB video file requires about 500MB of encoded space here. Based on the current pricing—0.0001 WAL per MiB per epoch (roughly two weeks)—the annual cost would be around tens to hundreds of dollars. Compared to some projects' permanent storage fees or other solutions with market volatility, this is indeed more stable and affordable.

However, there's a key prerequisite to keep in mind— the cost advantage is built on two fundamentals. One is the network scale—