Radar · Diagram Design
Four object-storage candidates scored 0–10 across five workload dimensions. The chart is a decision tool — the recommended option carries the only colored fill; everything else sits in the muted comparison palette.
RECOMMENDED
9 across the board, weakest only on large-object reads. The clear pick when small-file performance and Iceberg integration matter and you can run the operator yourself.
The reference implementation: top-class on large-object reads and write throughput, but a hard step down on operational simplicity once you factor in IAM, lifecycle, and bucket policy sprawl.
Capable across the board, but the lowest score on operational simplicity. Pick it when you already run Ceph or need block + object + file from one cluster — not for a fresh greenfield.
Solid middle path. No standout dimension — which is itself an answer when the team optimises for "boring, managed, fine."