Locally-Redundant Storage (LRS)
LRS replicates data three times within one data center located in a primary region. When LRS is enabled, Azure Storage only registers write requests as successful once data is written to three replicas. LRS provides at least 99.999999999% durability for objects during a given year.
LRS is offered at a low cost and can help you protect data against drive failure and server rack failure. However, LRS does not cover all data loss scenarios. Use LRS if it is less sensitive to loss, or if your data is restricted to one location due to compliance requirements.
Zone-Redundant Storage (ZRS)
ZRS performs replication across three Azure availability zones. Each Azure availability zone is an individual physical location with its own independent networking, power, and cooling. ZRS provides a minimum of 99.9999999999% durability for objects during a given year.
ZRS ensures data remains accessible even if an availability zone goes down. This is true both for write and read operations. Azure Storage only registers a write operation as successful, once it makes sure data is safely written to all three availability zones. You can use ZRS in a primary region to ensure consistency, high availability, and durability.
Geo-Redundant Storage (GRS)
GRS provides additional redundancy for data storage compared to LRS or ZRS. In addition to the three copies of data stored in one region, there are three copies stored in a paired Azure region. So GRS provides all the features of LRS storage in the primary zone, and additionally, provides a secondary LRS data storage in another region.
Because all read and write operations are still managed through one Azure data center, the read and write SLAs provided by Microsoft are the same as for ZRS and LRS data storage.
There are two disadvantages of GRS redundancy:
- Replication between regions is asynchronous and so data is propagated with a small delay
- The second region cannot be accessed or read until the storage account fails over
Read-Access Geo-Redundant (RA-GRS)
RA-GRS has all the same level of redundancy of standard GRS replication, with an additional benefit—the secondary copies stored in paired Azure regions are readable. This means that if your application is configured correctly, you can use multiple readable endpoints. This increases the SLA for read operations to 99.99%.
However, the SLA for write operations remains 99.9%, because a single area still controls write and update operations.
Due to their asynchronous replication, both types of GRS replication have some replication delay. You can use the LastSyncTime parameter to ensure you are reading the latest copy of the data.