Storage redundancy: LRS vs ZRS vs GRS vs GZRS
Verdict: Lowest cost, three copies in one datacentre, LRS. Across three zones in the region, ZRS. Copied to a distant second region, GRS. Zones in-region plus geo-replication, GZRS.
| Criterion | LRS | ZRS | GRS | GZRS |
|---|---|---|---|---|
| Copies | 3 in one datacentre | 3 across zones in-region | In-region plus a distant region | 3 zones plus a distant region |
| Survives | Rack or drive failure | A single zone outage | A regional disaster | Zone outage and regional disaster |
| Choose when | Cheapest, reproducible data | Stay writable through a zone outage in-region | Cross-region disaster recovery | Both zone and region resilience |
Rules
- Choose LRS for the lowest-cost redundancy: three replicas inside one datacentre, guarding rack and drive failures but exposed if that whole datacentre fails.
- Choose ZRS to stay available for reads and writes through a single zone outage while keeping all copies in the primary region.
- Choose GZRS to combine zone redundancy in the primary region with asynchronous replication to a geographically distant secondary region.
Traps
- LRS keeps all copies in one datacentre, so it offers no protection if that datacentre or zone fails; that needs ZRS.
- ZRS is primary-region-only and does not replicate to a secondary region, so it fails a geographic-distance requirement; that needs GZRS or GRS.