When configuring a RAID (Redundant Array of Independent Disks) system, one of the key considerations is whether to implement redundancy. Redundancy involves using extra disks to store duplicate copies of data, providing fault tolerance in case one disk fails. However, redundancy comes at the cost of reduced overall storage capacity. For some use cases where absolute storage space is a higher priority than data protection, a RAID configuration without redundancy may be preferred.
What is RAID?
RAID is a technology used to combine multiple physical disk drives into a single logical unit. RAID takes advantage of the parallelism of multiple disks to enhance performance and/or reliability compared to a single disk. There are several standard RAID levels, each with different mechanisms for distributing and replicating data across the array.
The main reasons to implement RAID include:
- Increased storage capacity – Combining multiple disks expands the total available storage space beyond the capacity of a single disk.
- Improved performance – Spreading I/O requests across multiple disks can increase throughput and reduce access times.
- Fault tolerance – Storing redundant data or parity information provides protection against disk failures.
Do I need redundancy in my RAID configuration?
The decision of whether to implement redundancy in a RAID configuration depends on your specific requirements and priorities:
- If absolute storage capacity is critical, then a RAID mode without redundancy may be preferable. Non-redundant RAID maximizes available space.
- If data protection is more important, choose a redundant RAID level. Redundancy comes at the cost of lower total capacity.
- Consider how critical the data is and the impact of potential disk failures. Higher redundancy may be warranted for mission critical data.
- Weigh the enhanced read performance of redundant RAID against the lower write performance and storage efficiency.
- Assess the cost trade-off between adding more disks for redundancy versus using that budget for more raw capacity instead.
In general, redundancy is recommended for most critical storage needs to protect against data loss. But scenarios focused exclusively on high capacity may warrant skipping redundancy.
Overview of non-redundant RAID levels
There are several standard RAID levels that provide no data redundancy. These include:
RAID 0
RAID 0, also known as disk striping, spreads data evenly across all disks in the array. RAID 0 provides improved performance by distributing the I/O load across multiple channels and drives. But it does not provide any fault tolerance. If one disk fails, all data in the array will be lost.
RAID 1
RAID 1, also known as disk mirroring, duplicates all data from one drive to a second drive. This protects data in case of a single disk failure, but cuts the total capacity in half. RAID 1 provides fault tolerance at the expense of storage efficiency.
RAID 5
RAID 5 stripes data and distributed parity information across all the disks. The parity allows the array to reconstruct data if one disk fails. But RAID 5 still requires a minimum of 3 disks, reducing overall capacity versus JBOD.
RAID 10
RAID 10 combines mirroring and striping by creating mirrored pairs of disks and then striping data across the pairs. This provides both fault tolerance and improved performance. Total array capacity is cut in half.
Non-redundant RAID performance
Non-redundant RAID levels can provide performance benefits compared to a single disk:
- Increased throughput – Distributing I/O across multiple disks allows more concurrent operations.
- Faster access – Data is spread across more physical media, reducing contention.
- Balanced load – All disks contribute to handling the workload.
However, redundant RAID levels provide more consistent performance by avoiding degradation when rebuilding failed disks. Overall, RAID 0 offers the best peak performance of non-redundant options.
When is non-redundant RAID acceptable?
There are some scenarios where the extra capacity of non-redundant RAID may outweigh the risks:
- Storing easily replaced data like caches or temporary files.
- Systems with robust, frequent backups making recovery reliable.
- Data warehouse/analytics workloads focused on capacity over availability.
- Non-critical consumer systems where cost trumps reliability.
However, for most business critical systems, redundant RAID or other fault tolerance measures are still recommended.
How reliable is non-redundant RAID?
Non-redundant RAID levels provide no protection against disk failures. A single disk failure will result in full array failure and complete data loss. Reliability is equivalent to the poorest performing individual disk:
- MTBF ratings are lower than multi-disk systems with redundancy.
- Failure of 5% of non-redundant RAID disks results in 5% array failure rate.
- Additional disks add more points of failure.
To improve reliability, deploy higher quality enterprise-class drives, implement hot spares, schedule frequent backups, and be prepared to immediately rebuild failed arrays.
Non-redundant RAID storage capacity
Total storage capacity is maximized without dedicating disks to redundancy. For an array with N disks:
- RAID 0 capacity equals sum of all disks.
- JBOD capacity equals sum of disks.
- RAID 1 capacity equals size of smallest disk.
RAID 0 provides the full cumulative capacity, but consider JBOD for better reliability and more flexible sizing.
Cost considerations for non-redundant RAID
Without redundancy, more disks mean lower cost per gigabyte of storage. But there are other financial factors:
- Higher likelihood of failure impacts maintenance and replacement costs.
- Lack of redundancy requires more frequent and comprehensive backups.
- Potential downtime and recovery costs are higher.
- Performance per dollar improves with more disks.
Calculate the total cost of ownership over the lifespan of the array, incorporating failure risks.
Best practices for implementation
To maximize performance and reliability of non-redundant RAID:
- Use higher speed enterprise SSDs rather than consumer HDDs.
- Configure arrays with at least 4-6 disks minimum to benefit from parallelism.
- Spread data evenly across all disks to balance load.
- Monitor disk health metrics and replace failing drives immediately.
- Maintain hot spares to allow quick rebuilding if disks fail.
Alternatives to consider
There are alternatives that provide redundancy while still offering large capacity:
- RAID 6 – Provides double distributed parity for additional fault tolerance.
- RAID-Z – ZFS proprietary implementation similar to RAID 5.
- Erasure Coding – More efficient distribution of parity than RAID 5/6.
- Distributed File Systems – GlusterFS, HDFS, Ceph replicate across nodes.
Also consider supplementing RAID with regular backups, snapshots, and replica copies to guard against multiple disk failures.
Software vs. hardware RAID
RAID can be implemented via dedicated hardware RAID controllers or via software in the operating system. There are pros and cons of each approach:
| Software RAID | Hardware RAID |
|---|---|
| Lower cost by using existing system resources | Dedicated RAID controller improves performance |
| Platform dependent, limited to OS support | Works independently of OS and processors |
| CPU load for RAID calculations | Offloads RAID processing overhead from main CPU |
| More flexibility in RAID levels | Typically supports limited RAID modes |
For most use cases, software RAID provides sufficient performance and flexibility at a lower cost. But hardware RAID may benefit performance critical applications.
How to recover data from failed non-redundant RAID
With non-redundant RAID, a single disk failure will result in full array failure. However, there are some recovery options:
- Restore from recent backups – Ensure backups are as frequent as data criticality warrants.
- Try rebuilding array from remaining disks – May work if failure was intermittent.
- Disk recovery services – Expensive, not guaranteed.
- Targeted disk image recovery – Recover only critical subsets of lost data.
Prevention is crucial with non-redundant RAID. Schedule regular backups, monitor disk health, hot swap failed drives quickly, and consider supplemental redundancy schemes.
Choosing a file system to pair with non-redundant RAID
The choice of file system to format non-redundant RAID arrays includes:
- EXT4 – Mature Linux file system with good performance.
- XFS – High performance file system for Linux.
- Btrfs – Includes snapshots and detection of silent data corruption.
- ZFS – Robust with native software RAID support.
- NTFS – Default modern Windows file system.
Consider file systems with checksums to detect corruption like ZFS and Btrfs. Schedule frequent scrubs to identify issues early.
Ideal applications for non-redundant RAID
Some examples of applications suitable for non-redundant RAID include:
- Scratch space – Temp storage for processing before archiving.
- Application caches – Transient data like web caches.
- Logging – Append-only logs. Tolerates some data loss.
- Media repositories – Easy to reacquire video, image, audio data.
- Big data analytics – Large data sets benefiting from capacity.
Any application where absolute storage capacity is the priority over resilience. Ensure adequate data recovery mechanisms are in place.
Conclusion
Non-redundant RAID levels like RAID 0 can maximize storage capacity by avoiding the space overhead of data protection. However, they come at the cost of increased risk of data loss in the event of drive failures. Weigh requirements carefully when evaluating redundant versus non-redundant RAID configurations. For mission critical data, redundant RAID or other fault tolerance schemes are still the safest choice.