Walrus's low-cost storage strategy is indeed eye-catching—subsidy period of 1TB for only $50 a year, non-subsidy monthly fee of $250. Compared to Filecoin, it's 80% to 100 times cheaper, and even hundreds of times cheaper than Arweave. It seems like a good deal, but once you delve into its epoch billing logic and actual settlement rules, you'll find many hidden traps in this pricing system.

I've been working on cost optimization in blockchain and cloud computing for many years, and I think it's necessary to break down these details to prevent everyone from being fooled by surface numbers and ending up paying more.

First, let's talk about this thing called epoch. Walrus divides time into epochs for settlement—testnet epochs are 24 hours, and the mainnet is supposedly designed for 14 days, but some documentation mentions it might still be 24 hours. This hasn't been finalized yet. All storage fees, node rewards, and committee rotations are calculated based on epochs. It sounds logical, but the real pitfalls are in the rules of "prepay by epoch" and "even if less than an epoch, billing still covers the entire epoch."

Here's a real-world example. Suppose you want to store a temporary file for only 3 days, but it happens to cross two epochs. If the mainnet epoch is truly 14 days, then storing 3 days of data would cost you for 28 days—more than nine times the cost. Even if the epoch is set to 1 day, 3 days of data might still require paying for 4 epochs (depending on exactly when you upload), which doubles the cost. What's more frustrating is that once the data is deleted, there is no refund mechanism.