Walrus is a project developed by an international team of developers. While its decentralized structure means it's not tied to a specific country, the team operates on a global scale. This approach ensures Walrus is an infrastructure solution that caters to the entire world, not just a particular region.
Problem Statement
In a nutshell a Complete Data Storage protocol allows a writer to write a blob to a network of storage nodes (Write Completeness), and then ensures that any reader can read it despite some failures and byzan- tine behaviour amongst storage nodes (Validity); and read it consistently, despite a potentially byzantine writer (Read Consistency). More formally:
Definition 1 (Complete Data Storage). Given a network of n = 3f + 1 nodes, where up to f are byzantine, let B be a blob that a writer W wants to store within the network, and share it with a set of readers R. A protocol for Complete Data Storage guarantees three properties:
1. Write Completeness: If a writer W is honest, then every honest node holding a commitment to blob B |B|
eventually holds a part p (derived from blob B), such that B can be recovered from O |p| parts.
2. Read Consistency: Two honest readers, R1 and R2, reading a successfully written blob B either both
succeed and return B or both return ⊥.
3. Validity: If an honest writer W successfully writes blob B, then an honest reader R holding a commitment
to B can successfully read B.
We present the ACDS protocols in a context where the storage node set is fixed and static. And in sub- sequent sections describing its use within Walrus, we discuss how it is used with changing committees of storage nodes.
3.2 The Red Stuff encoding protocol
In this section, we present the final design of Red Stuff by iterating first through two straw man designs and eliminate their inefficiencies.
3.2.1 1st straw man design: full replication
The simplest protocol uses full replication in the spirit of Filecoin [28] and Arweave [44]. The writer W broadcasts its blob B along with a binding commitment to B (e.g., HB = hash(B)), to all storage nodes and then waits to receive f + 1 receipt acknowledgments. These acknowledgments form an availability certificate which guarantees availability because at least one acknowledgement comes from an honest node. The writer W can publish this certificate on the blockchain, which ensures that it is visible to every other honest node, who can then request a Read(B) successfully. This achieves Write Completeness since eventually all honest nodes will hold blob B locally. The rest of the properties also hold trivially. Notice that the reader never reads ⊥.
Although the Full Replication protocol is simple, it requires the writer to send an O(n|B|) amount of data on the network which is also the total cost of storage. Additionally, if the network is asynchronous, it can cost up to f + 1 requests to guarantee a correct replica is contacted, which would lead to O(n|B|) cost per recovering storage node with a total cost of O(n2|B|) over the network. Similarly, even a read can be very efficient in asynchrony, as the reader might need to send f + 1 costing O(n|B|).
#walrus #WalrusProtocol #Binance #writetoearn
