An S3 API on your own hardware, with erasure coding and site replication that was actually tested
S3-compatible object storage (MinIO)
Where they started
Cloud object storage costs grew with customer retention requirements, and the product needed an on-premises deployment story for customers who could not use a public cloud.
The problem, precisely
The company had to operate object storage they had only ever consumed, including an erasure-coding layout and a replication policy they could defend to their own customers.
What we did
Built for two independent failure domains from the start, tested drive, node and site failure in sequence, and documented the capacity maths including the overhead they were paying.
Delivered
- Distributed deployment across two sites with an explicit erasure-coding overhead calculation
- Site replication configured with a measured RPO and a documented failover direction
- Object Lock and retention policy for regulatory data, tested against deletion attempts
- Failure tests: one drive, then one node, then one site, each measured
- Capacity and growth model including EC overhead, so the cost story holds up
Technologies used
Measured result
Every number below was measured and reported to the customer. Where we do not have a measured figure, the entry is a fact rather than a percentage — we would rather be precise than impressive.
- 480M objects under management
- 1 site loss survived with no data loss, tested
- — cost per terabyte per year, modelled against cloud
- 100% of retention rules enforced by Object Lock, not process
Erasure coding overhead is a first-class number, not a detail. If you cannot state it, you cannot price your own storage.