What Exactly Is Erasure Coding in Decentralized Storage?

Erasure coding in decentralized storage is how several major networks keep data recoverable after nodes fail, without paying the full cost of storing multiple complete copies. It's a real, well-documented technique with real math behind the savings, and it's genuinely different from replication, the copy-the-whole-thing approach. This article covers how erasure coding actually works, the concrete numbers real networks like Storj and Sia publish for it, and where it fits against the alternative. AIOZ Storage is not one of the networks using it, its own documentation describes a different mechanism, replication, and this article says so plainly rather than implying otherwise.

TL;DR:

• Erasure coding splits a file into n fragments where any k of them can reconstruct the original, tolerating the loss of up to n-k fragments without losing the file.
• It beats full replication badly on storage cost: Storj's own documentation notes that replicating data 10x still can't match the durability of a k=16, n=32 erasure code, which only costs 2x the storage.
• Storj and Sia both use erasure coding as their core, default redundancy mechanism, with real published numbers. AIOZ Storage's own documentation describes a different mechanism, replication, not erasure coding.

What Is Erasure Coding?

Erasure coding takes a block of data and encodes it into n total fragments, where any k of those n fragments, not a specific k, any k, are enough to reconstruct the original data. The two numbers, k and n, define the scheme entirely: k is the minimum needed to recover the file, n is the total number generated and spread across storage nodes. Lose up to n-k of the fragments to node failure, and the file is still fully recoverable. This is the same mathematical idea behind Reed-Solomon codes, the error-correcting technique used in CDs and DVDs to survive scratches, applied here to nodes going offline instead of physical damage to a disc.

The storage cost of a scheme is its expansion factor, n divided by k. A k=10, n=20 scheme has an expansion factor of 2, meaning a 10 MB file becomes 20 MB total spread across fragments, each roughly 1 MB. Push both numbers up proportionally, say to k=20, n=40, and the expansion factor stays the same while durability improves, since there are now more total fragments that could fail before the file becomes unrecoverable.

Why Erasure Coding Beats Full Replication on Cost

Full replication, storing complete, separate copies of a file across multiple nodes, is the simpler alternative, and it's meaningfully more expensive for the same durability. Storj's own documentation makes the comparison directly: "replicating data at 10x can't beat erasure codes with k=16, n=32, which is an expansion factor of only two." Ten complete copies of a file, a 10x storage cost, still doesn't match the fault tolerance of an erasure-coded scheme costing only twice the original file's size. The mechanism behind that gap is straightforward: replication only survives losing entire copies at a time, while erasure coding survives losing any combination of fragments up to its threshold, a much finer-grained and more efficient form of redundancy.

Real Networks Using Erasure Coding

Storj, a decentralized S3-compatible storage network, runs erasure coding as its core mechanism with published production parameters: k=29, n=80. A file needs any 29 of 80 fragments to be recoverable, and Storj's own documentation describes a repair process that kicks in once the number of healthy fragments drops to a repair threshold, regenerating lost pieces before the file's durability is actually at risk.

Sia, another decentralized storage network, uses Reed-Solomon erasure coding with a 10-of-30 scheme: files are striped into chunks, each chunk erasure-coded into 30 pieces, and any 10 of those 30 are enough to recover it. Sia's own documentation notes this tolerates up to 20 of 30 hosts going offline simultaneously without data loss.

Both are real, currently operating networks, and both numbers come directly from each project's own documentation, not estimates.

Where Filecoin Fits

Filecoin, another major decentralized storage network, supports erasure coding, but not as its automatic default. Filecoin's own documentation describes it as something a client requests when making a storage deal, specifying the redundancy scheme they want. Filecoin's primary, protocol-level approach to redundancy is actually replication across multiple independent storage providers, a client choosing to store several full copies with different providers rather than one erasure-coded set. Worth knowing before assuming every major decentralized storage network defaults to erasure coding the way Storj and Sia do, some default to replication instead and offer erasure coding as an option.

What AIOZ Storage Actually Uses

AIOZ Storage's own documentation states plainly that its underlying network "inherently supports replication by default," full copies of data spread across nodes, the same category of mechanism as Filecoin's default path, not the erasure-coding approach Storj and Sia run as their core design. AIOZ's documentation does not describe an erasure-coding scheme, doesn't publish a replication factor, and doesn't use the terms "erasure coding" or "parity" anywhere in its published material. If you're evaluating AIOZ Storage specifically, the cost-efficiency math above doesn't apply to it, don't assume it does. This article is general education about a real, differentiated technique other networks use, not a claim about how AIOZ Storage works.

Frequently Asked Questions

What does the "k" and "n" mean in erasure coding?
n is the total number of fragments a file gets split into. k is the minimum number of those fragments needed to reconstruct the original file. Any k of the n fragments work, not a specific subset.

How much more storage does erasure coding need compared to the original file?
Its expansion factor, n divided by k. A k=29, n=80 scheme (Storj's production parameters) has an expansion factor of about 2.76, meaning roughly 2.76x the original file's size in total fragment storage.

Is erasure coding always cheaper than replication?
For equivalent durability, yes, and often dramatically so. Storj's own documentation states that replicating data 10x still doesn't match the fault tolerance of an erasure code with only a 2x expansion factor.

Which decentralized storage networks actually use erasure coding?
Storj (k=29, n=80) and Sia (10-of-30) both run it as their core, default redundancy mechanism, per each project's own published documentation.

Does Filecoin use erasure coding?
It supports erasure coding as an option a client can request for a storage deal, but Filecoin's default, protocol-level redundancy approach is replication across multiple storage providers, not automatic erasure coding.

Does AIOZ Storage use erasure coding?
No. AIOZ Storage's own documentation describes replication, full copies of data across its network, as its redundancy mechanism, not erasure coding. AIOZ's material doesn't use the term "erasure coding" anywhere.

Is replication a worse approach than erasure coding?
Not necessarily worse, just a different tradeoff: simpler to reason about and implement, at a real storage-cost premium for the same durability level. Which one a network chooses depends on what it's optimizing for.

Related Reading

• What Is Data Replication? How Distributed Storage Systems Stay Available - the redundancy mechanism AIOZ Storage's own documentation actually describes, covered as its own topic with real numbers from HDFS and Cassandra.
• AIOZ Storage Explained: S3-Compatible Decentralized Storage - what AIOZ Storage is overall, and what its documentation does and doesn't confirm about redundancy.
• What Is Amazon S3? Object Storage Explained - the object storage model this article's redundancy discussion sits on top of.
• AIOZ Storage vs Storj: Two Decentralized S3-Compatible Platforms Compared - a direct comparison including the two networks' different redundancy approaches.
• Data Durability vs Availability: What 11 Nines Means - the two guarantees erasure coding and replication are both trying to earn, with AWS and Google Cloud's own published numbers.

References

• Storj Documentation: Understanding File Redundancy, Durability, Expansion Factors, and Erasure Codes
• Storj Documentation: Understanding File Repair in a Distributed Network
• Sia Documentation: Is My Data Secure?
• Filecoin: How Storage and Retrieval Deals Work