Erasure coding
A family of mathematical techniques that take k data symbols and produce n = k + m total symbols, such that any k of the n can reconstruct the original. Used for storage durability, network reliability, and (in SkyeConnex's case) sovereignty topology.
The core idea
Replication stores N independent copies. Erasure coding produces a mathematical encoding where fewer than the full set of chunks is sufficient for reconstruction. The trade-off: replication is operationally simpler; erasure coding is dramatically more storage-efficient at equivalent failure tolerance.
Storage economics
Triple-replication carries 3× storage overhead and tolerates loss of 2 of 3 copies. RS(5,2) — SkyeConnex's choice — carries 7/5 = 1.4× overhead and tolerates loss of 2 of 7 shards. Same fault tolerance; 53% less storage cost.
The security property replication doesn't give
With triple-replication, any single provider has a full readable copy of your data. A breach exposes everything. A subpoena to that one provider reveals everything. With RS(5,2), any single provider holds 2/7 of your data — mathematically, fewer than 5 of 7 shards reveal nothing about the original. The provider you trust least is no more dangerous than the provider you trust most. This is the property that makes erasure coding the right primitive for sovereignty, not just durability.
Common erasure codes
- Reed-Solomon — the family SkyeConnex uses. Optimal MDS codes; finite-field arithmetic; well-studied.
- Cauchy Reed-Solomon — a variant with XOR-only encoding.
- Local Reconstruction Codes (LRC) — used by Microsoft Azure for single-failure repairs without reading all shards.
- Fountain codes / LT codes — rateless codes for streaming and broadcast.
Erasure coding inside vs across
Many cloud providers use erasure coding internally — Backblaze B2 famously uses RS(17,3) within its data centres. This protects against drive failures within one provider. It does not give you sovereignty across providers. For sovereignty, the erasure-coding layer must operate across distinct legal entities in distinct jurisdictions. Read the full primer →
Engineering considerations
Frame size, parameter choice (k, m), shard distribution policy, and integrity verification all interact. SkyeConnex's RS(5,2) at 5 MB frame size is the sweet spot for sovereign multi-cloud storage: minimal overhead, strong fault tolerance, no single provider holds enough to read.
Related terms
See also
Posts that mention Erasure coding
Reed-Solomon erasure coding for storage: a practical primer
Reed-Solomon codes are decades old. They're suddenly central to sovereign cloud architecture. Here's what they do, the math you need, and th…
Read → Architecture · 7 min readWhy multi-cloud RAID beats multi-cloud sync
Most multi-cloud storage products replicate your file inside each provider. SkyeConnex erasure-codes it across them. That distinction is the…
Read →