RAID 5 is a popular disk or solid state drive (SSD) subsystem that increases safety by computing parity data and increases speed by interleaving data across multiple drives (PCMag, 2023). It requires a minimum of 3 drives and is able to withstand a single drive failure without losing data (Network Encyclopedia, 2023).
With RAID 5, data is striped across multiple drives, similar to RAID 0. However, unlike RAID 0, RAID 5 also computes and writes parity information to a dedicated drive in the array. The parity drive contains calculated error-correcting information that can be used to rebuild data in case of a single drive failure (PCMag, 2023).
This parity data provides redundancy and fault tolerance. If a single drive fails, the missing data can be recreated using the parity drive. This provides improved reliability over a single drive or RAID 0 array, while also providing better performance than a mirrored RAID 1 array (Network Encyclopedia, 2023).
Minimum Number of Drives
RAID 5 requires a minimum of 3 drives to configure an array. This is because RAID 5 uses distributed parity, which means that it stripes data across multiple drives while dedicating one drive’s worth of capacity for parity information. The parity information is distributed evenly across all drives and can be used to reconstruct data in the event of a single drive failure.
According to drivesaversdatarecovery.com, “Even though the minimum drives for RAID 5 is three, most users opt for four drives because of speed, fault tolerance and storage capacity.” With only 3 drives, the array has no fault tolerance if another drive fails during rebuild after a first drive failure. Four drives or more are recommended for fault tolerance.
Drive Failure Tolerance
One of the key benefits of RAID 5 is its ability to tolerate the failure of one drive without data loss. This is achieved through a distributed parity scheme that stores parity information across all the drives in the array https://discussions.apple.com/thread/1953516. If any single drive fails, the missing data can be recalculated from the parity information on the remaining drives. This provides fault tolerance and avoids downtime from having to restore from backups.
RAID 5 can continue operating normally with one failed drive. However, if a second drive fails before the failed drive is replaced, data loss will occur. Therefore, it is crucial to replace failed drives promptly to regain fault tolerance. Overall, the ability to withstand a single drive failure makes RAID 5 a popular choice for providing redundancy without sacrificing too much usable capacity.
Recommended Number of Drives
While RAID 5 can technically be created with just 3 drives, most experts recommend using 4-6 drives for optimal performance and redundancy. According to Seagate’s RAID calculator, using 4-6 drives provides the best balance between storage efficiency, performance, and fault tolerance for most RAID 5 implementations.
With just 3 drives, RAID 5 can only withstand a single drive failure before complete data loss. Adding a 4th or 5th drive increases redundancy and allows the array to survive multiple drive failures. The Seagate RAID calculator cites 4-6 drives as the “recommended number for performance” with RAID 5.
Additionally, some sources note that RAID 5 performance suffers with an odd number of drives. ServerFault explains that RAID 5 write performance is optimized with drives counts that are powers of 2 (4, 8, 16, etc.) due to more efficient stripe distributions. So 4 or 6 drives is preferable over 5 for performance.
While more than 6 drives can be used in RAID 5, most experts advise against it due to performance declines and high rebuild times when attempting to reconstruct a failed drive’s data across many remaining drives.
Performance Impact
RAID 5 provides good performance for read operations, since data can be read in parallel from multiple drives. However, write operations are slower compared to RAID 0 or RAID 1 due to the parity calculations that need to be performed on each write. According to this source, RAID 5 can achieve over 80% of the read performance of RAID 0, but write performance may be only 50-60% of RAID 0. The performance degradation for writes gets progressively worse as the array size increases.
Writes in RAID 5 require the following steps: 1) Read old data and old parity, 2) Calculate new parity, 3) Write new data and parity. Because of this process, the larger the array, the greater the write penalty. For arrays above 6-8 drives, the RAID 5 write penalty becomes quite high and alternatives like RAID 10 are recommended instead.
RAID 5 Arrays Larger than 6 Drives
RAID 5 arrays with more than 6 drives come with some significant drawbacks related to rebuild times and performance. As the array size grows, rebuild times increase exponentially. According to this discussion, large RAID 5 arrays are extremely risky due to the long rebuild times. If another drive fails during a rebuild, the entire array will be lost. Rebuilding a 10+ TB RAID 5 array can take multiple days, as noted here.
In addition to long rebuild times, large RAID 5 arrays suffer performance penalties during rebuilds and when dealing with multiple disk failures. As the array size grows, the performance impact becomes more severe. For these reasons, experts recommend avoiding RAID 5 arrays with more than 6 drives. Arrays of that size introduce too much risk and performance instability.
Better Alternatives for Large Arrays
For arrays with a large number of drives (generally 6 or more), RAID 5 becomes less ideal. The larger the array, the higher the risk of multiple drive failures occurring before a failed drive can be rebuilt. In these cases, RAID 6 or RAID 10 are usually better choices.
RAID 6 is similar to RAID 5 but provides double distributed parity. This means the array can withstand the failure of up to two drives without data loss. RAID 6 requires a minimum of 4 drives. The tradeoff is reduced usable capacity compared to RAID 5. A 4 drive RAID 6 array would only have the usable capacity of 2 drives.
RAID 10 provides performance and redundancy by mirroring two drives into one set then striping data across multiple sets. This allows for high throughput while also providing fault tolerance. At least 4 drives are required for RAID 10. Usable capacity is equal to half the total capacity when using two drive mirror sets. RAID 10 is a good choice for applications requiring high performance and redundancy with a limited number of drives.
For large arrays where performance and redundancy are top priorities, most experts recommend RAID 6 or RAID 10 over RAID 5. The capacity overhead is justified by the added protection against multiple drive failures. RAID 5 should be limited to smaller arrays where a single disk fault tolerance is sufficient.
(Sources: Reddit, Spiceworks)
Drive Size Considerations
Drive size is an important factor to consider when configuring RAID 5 arrays. As drive sizes continue to increase, rebuild times also increase proportionally. This is because rebuilding the array requires reading all the data from the remaining disks in order to reconstruct the lost data from the failed drive. With larger drives, there is more data to read which increases rebuild times.
For example, rebuilding a 1TB drive may take around 1 hour, but a 6TB drive in the same array could take 6 hours to rebuild. Long rebuild times increase the risk of a second drive failing before the rebuild completes, which would result in data loss.
To mitigate this risk, larger drives may warrant using RAID 6 instead of RAID 5 to provide double parity and tolerance for up to two drive failures. RAID 6 helps ensure the array can survive a second drive failure during a prolonged rebuild with larger drives. Or, using a greater number of smaller drives in the RAID 5 array can reduce average rebuild times.
When configuring RAID 5, carefully consider the drive sizes involved and the impact on rebuild times. Larger drives mean longer rebuilds, so plan accordingly with more parity, smaller drives, or stronger fault tolerance.
Sources:
[1] https://jmetz.com/2022/10/raid-5-recommendations/
When to Use RAID 5
RAID 5 is a good option for arrays with fewer than 6 drives that need moderate performance and don’t house critical data. The parity scheme in RAID 5 allows for 1 drive failure tolerance while still providing improved read speeds over a single drive. However, RAID 5 is not recommended for large arrays or mission critical applications due to the increased risk of data loss during rebuilds.
For arrays with 5 or fewer drives, RAID 5 provides a nice balance of storage efficiency, performance, and redundancy. During normal operation, read speeds will be faster than a single drive since data can be striped across multiple disks. Write speeds will be slower than RAID 0 or 10 due to the parity calculation overhead, but still faster than a single drive. If a single drive fails, the array will remain operational thanks to the distributed parity, allowing you time to replace the failed drive.
RAID 5 is a good fit for use cases like home media servers where some redundancy is desired, but maximum performance is not critical. The lower cost compared to RAID 10 also makes RAID 5 attractive for small arrays on a budget. Just be mindful that RAID 5 is not foolproof – an unrecoverable read error during a rebuild can result in data loss. So avoid RAID 5 for storing data you cannot afford to lose.
Summary
To summarize the key points about the number of drives required for RAID 5:
- The minimum number of drives required is 3.
- RAID 5 can tolerate 1 drive failure without data loss.
- Most implementations recommend at least 5 drives to balance performance and redundancy.
- Performance degrades significantly in large arrays over 6-8 drives.
- RAID 6 or RAID 10 are better alternatives for arrays larger than 6-8 drives.
- Larger capacity drives can reduce the drive count for a given storage size.
- RAID 5 is best suited for medium sized arrays that need redundancy but don’t require peak performance.
The bottom line is that the recommended number of disks for RAID 5 is between 3 and 8, with 5-6 drives being the ideal balance for most use cases.