This chapter covers Azure Storage Redundancy Options, focusing on Locally Redundant Storage (LRS) and Geo-Redundant Storage (GRS). Understanding these redundancy levels is critical for the AZ-900 exam, as it tests your ability to choose the right storage durability based on business requirements, cost, and compliance. This objective area carries approximately 15-20% of the exam weight, making it a high-priority topic. By the end of this chapter, you will be able to explain the differences between LRS, GRS, and their variants, and select the appropriate option for a given scenario.
Jump to a section
Imagine you run a small business and store your daily cash in a single bank vault (your primary data center). This vault is your only copy of the money. If a fire or flood destroys that vault, your cash is gone forever. That's like Locally Redundant Storage (LRS) — your data is replicated three times within a single physical location, but if that location is destroyed, you lose all copies. Now consider a more sophisticated approach: your bank vault automatically sends a sealed, tamper-proof copy of every day's cash to a second vault in a different city every night. If your primary vault burns down, you can retrieve your cash from the secondary vault. This is Geo-Redundant Storage (GRS). The mechanism is not just a simple copy; the data is written to the primary region first, and then asynchronously replicated to a paired secondary region hundreds of miles away. If a regional disaster occurs, Microsoft initiates a failover to the secondary region, and you can read your data. However, there is a catch: the replication is asynchronous, meaning there is a small window (typically 15 minutes) during which recent writes might not yet be replicated to the secondary region. So, like the bank's nightly transfer, you could lose the last few transactions if the disaster strikes before the next sync. This is the trade-off between cost and durability.
What is Azure Storage Redundancy and Why Does It Matter?
Azure Storage is a core service that stores data such as blobs, files, queues, and tables. When you store data in Azure, it is physically stored on multiple disks across multiple servers to protect against hardware failures. However, hardware failures are not the only risk. A data center could be taken offline by a natural disaster, a power outage, or even a software bug that corrupts data. To address these risks, Azure offers several redundancy options that determine how many copies of your data are stored and where they are located.
Redundancy is about durability (the likelihood that data will survive failures) and availability (the ability to access data when needed). The trade-off is cost: higher durability and availability come at a higher price. For the AZ-900 exam, you must understand four primary redundancy options: LRS, ZRS, GRS, and GZRS. This chapter focuses on LRS and GRS, but we will briefly contrast them with ZRS and GZRS.
How Azure Storage Replication Works – The Mechanism
When you create a storage account, you choose a redundancy option. Azure automatically handles all replication behind the scenes. The process varies by option:
LRS (Locally Redundant Storage): Your data is replicated three times within a single data center in the primary region. All three replicas are in the same availability zone (if the region supports zones) or across different fault domains and update domains within the same data center. If a disk fails, Azure automatically reads from another replica. However, if the entire data center is destroyed, all copies are lost.
GRS (Geo-Redundant Storage): Your data is first replicated three times within the primary region (like LRS) and then asynchronously replicated to a paired secondary region hundreds of miles away. In the secondary region, the data is also stored three times (LRS within that region). The replication is asynchronous, meaning there is a lag (typically less than 15 minutes, but no SLA). During a regional disaster, Microsoft will initiate a failover to the secondary region, and you can then read your data (write access is not enabled by default unless you enable read-access geo-redundant storage, RA-GRS).
ZRS (Zone-Redundant Storage): Your data is replicated synchronously across three Azure availability zones within the primary region. Each zone is a physically separate data center with independent power, cooling, and networking. This protects against a single zone failure but not a region-wide disaster.
GZRS (Geo-Zone-Redundant Storage): Combines ZRS in the primary region with asynchronous replication to a secondary region. This offers the highest durability.
Key Components and Tiers
Storage Account Types: The redundancy option is selected when creating a storage account. The type of storage account (general-purpose v2, blob storage, etc.) does not change the redundancy behavior.
Primary and Secondary Regions: Each Azure region is paired with another region (e.g., East US pairs with West US). These pairs are at least 300 miles apart. GRS and GZRS use these pairs.
Failover: For GRS and GZRS, you can either rely on Microsoft-managed failover (initiated by Microsoft in the event of a disaster) or you can perform a customer-managed failover (for certain storage account types). After failover, the secondary region becomes the new primary. Note that failover is a one-way process; you cannot fail back automatically.
Read Access: Standard GRS does not allow reading from the secondary region unless you enable RA-GRS (Read-Access Geo-Redundant Storage). RA-GRS is a variant that provides read-only access to the secondary region at all times, not just during failover.
Durability and SLA: LRS offers 11 nines (99.999999999%) durability over a given year for objects in a storage account. GRS offers 16 nines (99.99999999999999%) durability. The availability SLA for LRS is 99.99% for read and write operations; for RA-GRS it is 99.99% for reads and 99.9% for writes (because writes are still only on primary).
Comparison with On-Premises Equivalent
In an on-premises environment, redundancy is typically achieved through RAID (Redundant Array of Independent Disks) or backup tapes. RAID 1 mirrors data to two drives, similar to LRS but at a smaller scale. Offsite backups are analogous to GRS, but often require manual processes and are not synchronous. Azure automates replication and failover, reducing administrative overhead. The key difference is that Azure's redundancy is managed as a service; you do not need to provision hardware or manage replication schedules.
Azure Portal and CLI Touchpoints
In the Azure portal, you configure redundancy when creating a storage account under the "Basics" tab, in the "Redundancy" dropdown. You can also change redundancy later, but some changes may require data movement and incur downtime. Using Azure CLI, you can create a storage account with LRS using:
az storage account create --name mystorageaccount --resource-group myResourceGroup --location eastus --sku Standard_LRSFor GRS, change --sku to Standard_GRS. Other SKU values include Standard_RAGRS, Standard_ZRS, and Standard_GZRS.
You can also check the current redundancy setting with:
az storage account show --name mystorageaccount --resource-group myResourceGroup --query "sku.name"Business Scenarios
LRS Scenario: A development team needs a cost-effective storage for temporary test data. They choose LRS because losing the data is acceptable if a disaster occurs. Cost is low, and performance is good.
GRS Scenario: A financial institution stores transaction logs that must survive a regional disaster. They choose GRS to ensure data durability. They may also enable RA-GRS for read-only access during normal operations for compliance reporting.
GZRS Scenario: An e-commerce platform requires high availability and durability. They choose GZRS to survive both a zone failure and a regional disaster.
Pricing Models
LRS: Lowest cost. You pay for the storage capacity used, plus transaction costs.
GRS: Higher cost because of data replication to a secondary region and network egress charges.
RA-GRS: Slightly higher than GRS due to read access to secondary.
ZRS and GZRS: More expensive than LRS and GRS respectively because of synchronous replication across zones.
Important Limits and Defaults
The default redundancy for new storage accounts is LRS (though this may change; always verify).
Changing from LRS to GRS is allowed but requires copying data to a new account or using live migration (which may have constraints).
Data in the secondary region is not accessible until failover unless RA-GRS is enabled.
Failover can take hours, and data loss is possible (up to the replication lag).
Trap Patterns for the Exam
Common wrong answers include: - "LRS replicates data across multiple data centers" – Incorrect; LRS is within one data center. - "GRS provides synchronous replication" – Incorrect; it is asynchronous. - "RA-GRS allows writes to the secondary region" – Incorrect; it is read-only. - "ZRS replicates across regions" – Incorrect; ZRS is within one region across zones.
The exam loves to test the exact number of copies and locations. Remember: LRS = 3 copies in one data center; ZRS = 3 copies across 3 zones; GRS = 6 copies total (3 in primary, 3 in secondary); GZRS = 6 copies (3 across zones in primary, 3 in secondary).
Summary of Key Points
LRS: 3 copies, single data center, 11 nines durability, 99.99% availability.
GRS: 6 copies, two regions, 16 nines durability, 99.99% read (with RA-GRS) / 99.9% write.
ZRS: 3 copies across zones, single region, 12 nines durability, 99.99% availability.
GZRS: 6 copies across zones and regions, 16 nines durability.
Replication is automatic; no manual intervention needed.
Failover options: Microsoft-managed or customer-managed (for certain accounts).
Choose a redundancy option
When creating a storage account in the Azure portal, you select a redundancy option from the dropdown. This choice determines the durability, availability, and cost. For beginners, LRS is the default and cheapest. The exam expects you to know the trade-offs: LRS is suitable for non-critical data that can be regenerated, while GRS is for data that must survive a regional disaster. After selection, Azure automatically configures replication. You cannot change the option later without data movement in some cases, so choose carefully.
Data is written to primary region
Your application writes data to the storage endpoint in the primary region. For LRS, Azure immediately writes three copies within the same data center. For GRS, the write is first committed to the primary region (with LRS-like replication), then an asynchronous background process copies the data to the secondary region. The synchronous nature of the primary write ensures low latency for your application. The asynchronous geo-replication runs continuously, typically with a lag of less than 15 minutes.
Replication to secondary region (GRS only)
For GRS, after the primary write is confirmed, Azure queues the data for replication to the paired secondary region. The replication is block-level, meaning only changed blocks are sent, not the entire file. This minimizes network traffic. The secondary region stores the data three times (LRS within that region). You have no control over which secondary region is used; it is predetermined by Azure region pairing. The exam may ask about paired regions (e.g., East US pairs with West US).
Enable read access (optional for GRS)
By default, GRS does not allow reading from the secondary region. To enable read access, you must choose RA-GRS (Read-Access Geo-Redundant Storage) at account creation. With RA-GRS, you can read data from the secondary region at any time, but writes still go only to the primary. This is useful for scenarios where you need low-latency reads from a geographically distant location or for compliance. The exam tests that RA-GRS is read-only; never write.
Failover during disaster
If a primary region disaster occurs, Microsoft may initiate a failover to the secondary region. Alternatively, you can perform a customer-managed failover for certain storage account types. During failover, the secondary region becomes the new primary, and DNS records are updated. After failover, you can read and write to the new primary. However, data written just before the disaster that was not yet replicated is lost (the RPO is the replication lag). The exam expects you to know that failover is not instantaneous and data loss is possible.
Scenario 1: A Healthcare Startup Storing Patient Records
A healthcare startup deploys a patient records system on Azure. They need to comply with HIPAA regulations, which require data to be protected against regional disasters. They choose GRS for their storage account. The team configures the storage account with RA-GRS to allow read-only access to the secondary region for reporting purposes. During normal operations, writes go to East US, and reads can be served from West US for users on the west coast. If a hurricane hits East US, Microsoft fails over to West US, and the application continues with minimal downtime. The cost is higher than LRS, but the business accepts it for compliance. A common mistake would be choosing LRS to save costs, which would violate HIPAA and risk data loss.
Scenario 2: A Gaming Company Storing Player Profiles
A gaming company stores player profiles and game state data. They prioritize low latency and high availability. They choose LRS for each region where they have a presence, but they also implement application-level replication to another region. This gives them control over failover and avoids the cost of GRS. However, if a data center goes down, they lose data for that region until the application replicates. This is acceptable because player data can be regenerated from logs. The exam might ask which redundancy option is appropriate for non-critical data: LRS.
Scenario 3: A Financial Services Firm with Strict Durability Requirements
A financial services firm stores transaction logs that must be preserved for 7 years. They cannot afford any data loss. They choose GZRS (Geo-Zone-Redundant Storage) to protect against both a zone failure and a regional disaster. They also enable RA-GRS for read access. The cost is high, but the durability is 16 nines. They test failover quarterly using customer-managed failover. A common misconfiguration is using LRS, which would result in data loss if the data center is destroyed. The exam tests that GZRS is the most durable option.
Exactly What AZ-900 Tests on This Objective (Code: 2.4)
The objective "Describe storage redundancy options" is part of the larger domain "Azure Architecture and Services". The exam expects you to:
Differentiate between LRS, ZRS, GRS, and GZRS.
Know the number of copies and their locations (e.g., LRS: 3 copies in one data center; GRS: 6 copies across two regions).
Understand the durability and availability SLA differences.
Identify scenarios where each option is appropriate.
Know that RA-GRS provides read-only access to the secondary region.
Understand that GRS replication is asynchronous and has a recovery point objective (RPO) of typically less than 15 minutes.
Common Wrong Answers and Why Candidates Choose Them
"LRS replicates data across multiple data centers." Candidates confuse LRS with ZRS or GRS. LRS is only within one data center.
"GRS provides synchronous replication." Candidates think "geo" implies immediate, but it's asynchronous.
"RA-GRS allows writes to the secondary region." The 'R' stands for read-only, but candidates assume read/write.
"ZRS replicates data to a secondary region." ZRS is zone-redundant within one region, not geo.
"All redundancy options offer 16 nines durability." Only GRS/GZRS offer 16 nines; LRS offers 11, ZRS offers 12.
Specific Terms and Values That Appear Verbatim
11 nines (99.999999999%) for LRS durability.
16 nines (99.99999999999999%) for GRS durability.
99.99% availability SLA for LRS and RA-GRS reads.
99.9% availability SLA for RA-GRS writes (since writes are only on primary).
RPO of less than 15 minutes for GRS.
Paired regions are at least 300 miles apart.
Edge Cases and Tricky Distinctions
Changing redundancy: You cannot change from LRS to GRS without data movement unless you use live migration (which is not always available). The exam might present a scenario where you need to change and ask if it's possible.
Failover types: Know the difference between Microsoft-managed and customer-managed failover. Customer-managed is available for certain storage account types (e.g., general-purpose v2).
Data loss: Even with GRS, there is a possibility of data loss due to asynchronous replication. The exam may ask about the maximum data loss window.
Performance: LRS generally has lower latency than GRS because there is no geo-replication overhead.
Memory Trick for Eliminating Wrong Answers
Use the acronym L-Z-G-G in order of increasing durability and cost: - L = Local (one data center) - Z = Zone (across zones within region) - G = Geo (across regions) - G = Geo-Zone (both)
For any question about "most durable" or "highest availability," pick GZRS. For "lowest cost," pick LRS. If the question mentions "read access from secondary region," look for RA-GRS or RA-GZRS. If it says "asynchronous replication," it's GRS or GZRS. If it says "synchronous," it's LRS or ZRS.
Decision Tree
Does the data need to survive a regional disaster? If yes → GRS or GZRS. If no → LRS or ZRS.
Does the data need to survive a zone failure? If yes → ZRS or GZRS.
Is cost the primary concern? If yes → LRS.
Is read access from secondary needed during normal operations? If yes → RA-GRS or RA-GZRS.
LRS: 3 copies in one data center, 11 nines durability, 99.99% availability, lowest cost.
GRS: 6 copies across two regions, 16 nines durability, asynchronous replication (<15 min RPO).
RA-GRS: Allows read-only access to secondary region at all times.
ZRS: 3 copies across availability zones in one region, 12 nines durability.
GZRS: Combines ZRS and GRS for highest durability (16 nines).
Failover: Microsoft-managed or customer-managed; data loss possible up to RPO.
Durability values: LRS 11 nines, ZRS 12 nines, GRS/GZRS 16 nines.
Availability SLA: LRS 99.99% read/write; RA-GRS 99.99% read, 99.9% write.
These come up on the exam all the time. Here's how to tell them apart.
LRS (Locally Redundant Storage)
3 copies within a single data center
11 nines durability
Lowest cost
No protection against data center failure
Available in all storage account types
GRS (Geo-Redundant Storage)
6 copies across two regions (3 primary, 3 secondary)
16 nines durability
Higher cost due to geo-replication and egress
Protects against regional disaster
Asynchronous replication with <15 min RPO
Mistake
LRS replicates data to multiple data centers within the same region.
Correct
LRS replicates data within a single data center in the primary region. It does not use availability zones or other data centers.
Mistake
GRS provides synchronous replication, so no data is ever lost.
Correct
GRS replication is asynchronous. There is a lag (RPO < 15 minutes) during which recent writes may not be replicated. Data loss is possible during a disaster.
Mistake
RA-GRS allows writes to the secondary region.
Correct
RA-GRS provides read-only access to the secondary region. Writes are only accepted in the primary region.
Mistake
ZRS replicates data across multiple Azure regions.
Correct
ZRS replicates data across availability zones within a single region. It does not protect against a region-wide disaster.
Mistake
All storage redundancy options offer the same durability (99.9999999999%).
Correct
Durability varies: LRS: 11 nines, ZRS: 12 nines, GRS/GZRS: 16 nines.
LRS (Locally Redundant Storage) replicates your data three times within a single data center in the primary region. It protects against disk failures but not a data center disaster. GRS (Geo-Redundant Storage) replicates data three times in the primary region (like LRS) and then asynchronously replicates it to a paired secondary region, where it is also stored three times. GRS protects against a regional disaster, but has a recovery point objective (RPO) of less than 15 minutes, meaning data written just before a disaster may be lost. The exam expects you to know that LRS is cheaper and has lower durability (11 nines) compared to GRS (16 nines).
By default, no. Standard GRS does not allow access to the secondary region until a failover occurs. To enable read access during normal operations, you must use RA-GRS (Read-Access Geo-Redundant Storage). RA-GRS provides read-only access to the secondary region. This is useful for scenarios where you want to serve read requests from a geographically closer location or for compliance. The exam often asks whether RA-GRS allows writes — the answer is no, it is read-only.
LRS offers 11 nines of durability (99.999999999%), meaning that over a year, the expected risk of losing an object is about 0.000000001%. GRS offers 16 nines (99.99999999999999%), which is significantly higher. The difference is due to the geo-replication: if the primary region is destroyed, the secondary region still has your data. For the exam, memorize these numbers: LRS = 11, ZRS = 12, GRS = 16, GZRS = 16.
LRS has an availability SLA of 99.99% for both read and write operations. RA-GRS has 99.99% availability for reads (since reads can be served from secondary) and 99.9% availability for writes (since writes are only on primary). The exam may ask you to compare these numbers. Note that standard GRS (without read access) has the same write availability as LRS (99.99%) but no read SLA for secondary during normal operations.
You can change the redundancy option in the Azure portal under the Configuration blade, but not all changes are allowed. For example, you can change from LRS to GRS, but this may trigger a live migration that can take time and may have constraints. Some changes, like going from GRS to LRS, may require data movement to a new account. The exam might test that you cannot change from LRS to ZRS or GZRS without creating a new account. Always check the current limitations in the documentation.
GRS (Geo-Redundant Storage) replicates data three times in the primary region (within one data center) and then asynchronously to a secondary region. GZRS (Geo-Zone-Redundant Storage) replicates data synchronously across three availability zones in the primary region (like ZRS) and then asynchronously to a secondary region. GZRS offers higher durability because it protects against both a zone failure and a regional disaster. GZRS is more expensive than GRS. The exam expects you to know that GZRS is the most durable option.
During a regional disaster, Microsoft may initiate a failover to the secondary region. Alternatively, you can perform a customer-managed failover for eligible storage accounts. After failover, the secondary region becomes the new primary, and DNS records are updated. You can then read and write to the new primary. However, any data that was not replicated before the disaster is lost (the RPO window). Failover is not instantaneous and can take hours. After failover, you cannot automatically fail back; you must set up replication again. The exam may ask about data loss during failover.
You've just covered Azure Storage Redundancy Options (LRS/GRS) — now see how well it sticks with free AZ-900 practice questions. Full explanations included, no account needed.
Done with this chapter?