RAID (Redundant Array of Independent Disks) is one of the most misunderstood topics in server storage. It appears frequently in hosting specifications without explanation, and the most common misunderstanding, that RAID replaces backup, leads to data loss in situations where the configuration provides no protection.
InMotion Hosting dedicated servers use mdadm software RAID 1 (mirroring) across dual NVMe drives as the default configuration. This article explains what that means, what it protects against, what it does not protect against, and when different RAID configurations make sense for different workloads.
RAID Fundamentals
What RAID Does and Does Not Do
RAID distributes data across multiple physical drives to achieve one or both of two goals: performance improvement through parallelism, and fault tolerance through redundancy. The level of RAID determines which goal is prioritized.
What RAID does not do: protect against accidental deletion, software corruption, ransomware, or hardware failures that affect multiple drives simultaneously (fire, flood, or a power surge that damages both drives). These failure modes require backup, not RAID.
RAID Levels on Dual NVMe Drives
RAID 0: Striping
RAID 0 splits data across both drives in alternating blocks. A 100MB write becomes 50MB to Drive 1 and 50MB to Drive 2 simultaneously, completing in roughly half the time of a single-drive write.
Usable capacity: Full combined capacity (7.68TB on dual 3.84TB drives).
Read performance: Up to 2x sequential read throughput.
Write performance: Up to 2x sequential write throughput.
Redundancy: None. A single drive failure destroys all data on the array.
RAID 0 is appropriate for scratch storage, render caches, and non-critical temporary data where maximum throughput matters and the data can be regenerated. It is not appropriate for production databases, application data, or any data that cannot be reconstructed from an external source.
RAID 1: Mirroring (InMotion Hosting Default)
RAID 1 writes identical data to both drives simultaneously. Both drives contain a complete copy of all data. If one drive fails, the array continues operating from the surviving drive with no data loss.
Usable capacity: 50% of combined capacity (3.84TB on dual 3.84TB drives).
Read performance: Can read from either drive; software RAID can distribute reads for modest improvement.
Write performance: Must write to both drives; write throughput limited to single-drive write speed.
Redundancy: Survives one complete drive failure with no data loss.
RAID 1 is InMotion Hosting’s default configuration for dedicated servers with dual NVMe drives. For production databases, application data, and any workload where data loss is unacceptable, RAID 1 provides the right baseline protection. The 50% capacity trade-off is the cost of redundancy.
RAID 10: Striped Mirrors
RAID 10 requires four or more drives: drives are paired into RAID 1 mirrors, then those mirrors are striped together in RAID 0. This combines the performance of striping with the redundancy of mirroring.
Usable capacity: 50% of total drive capacity across all drives.
Read performance: 2x sequential read throughput (stripe across mirror pairs).
Write performance: Matches single-drive write speed (must write to mirror pairs).
Redundancy: Survives one drive failure per mirror pair; can survive multiple failures if they occur in different pairs.
RAID 10 on InMotion Hosting servers would require four NVMe drives, which is not the standard dual-drive configuration. For workloads requiring both maximum throughput and redundancy, a multi-server architecture with application-level replication (database primary/replica) often provides better results than a single server with four drives.
Software RAID vs. Hardware RAID
How InMotion’s mdadm RAID Works
InMotion Hosting uses mdadm (Multiple Device Administration), the Linux kernel’s software RAID implementation. This is a critical distinction from hardware RAID controllers, which use a dedicated processor on a RAID controller card to manage array operations.
Software RAID operations (parity calculation for RAID 5/6, mirroring writes for RAID 1) run on the server’s main CPU. On modern multi-core processors, this overhead is minimal for RAID 1: a RAID 1 mirror write requires no parity calculation, just writing to two devices simultaneously. The CPU overhead for mdadm RAID 1 on an NVMe array is typically under 1% on a 16-core AMD EPYC processor.
Advantages of Software RAID
No controller failure mode: Hardware RAID controllers can fail. When a proprietary RAID controller fails, the array is typically unreadable without an identical replacement controller. mdadm arrays can be read on any Linux system with the same version of mdadm.
No battery-backed write cache requirement: Hardware RAID controllers use battery-backed write cache to safely delay writes to disk. This cache is a failure point. mdadm RAID 1 writes directly to NVMe, which has built-in power-loss protection (PLP) on enterprise-grade NVMe drives.
Portability: An mdadm RAID 1 array can be moved to another server and reassembled. Drive metadata makes reassembly automatic.
Hardware RAID Advantages (and Why They Matter Less on NVMe)
Hardware RAID controllers historically provided two advantages over software RAID: battery-backed write cache for safe write acceleration, and dedicated processing to avoid CPU overhead on complex RAID levels (RAID 5, RAID 6).
NVMe drives with power-loss protection (enterprise NVMe, which InMotion uses) have built-in capacitors that flush write buffers to non-volatile storage on power loss. This eliminates the primary safety concern that battery-backed RAID cache addressed. And the CPU overhead argument was relevant when servers ran single-core or dual-core processors handling large parity calculations. On a 16-core EPYC doing RAID 1, the overhead is negligible.
NVMe RAID Performance Characteristics
RAID 1 sequential read performance can be slightly higher than a single drive if the software RAID driver distributes consecutive reads across both drives. In practice, mdadm RAID 1 read performance is approximately single-drive speed for sequential workloads and slightly better for random reads under concurrent access.
The write performance of RAID 1 matches single-drive write speed because both drives must receive the write before it is considered complete. On NVMe drives rated at 4GB/s sequential write, RAID 1 write throughput is approximately 4GB/s. This is fast enough for virtually any single-server workload.
Drive Failure and Rebuild Process
What Happens When a Drive Fails
When one drive in an mdadm RAID 1 array fails, the array continues operating in a degraded state from the surviving drive. Performance may drop slightly during degraded operation as all reads now come from a single drive, but the server stays online and data remains intact.
InMotion Hosting’s monitoring detects drive failures and initiates hardware replacement. Once the replacement drive is installed, mdadm rebuilds the array by copying all data from the surviving drive to the new drive.
Rebuild Times on NVMe
NVMe rebuild times are substantially faster than SATA SSD or spinning disk:
Spinning disk RAID 1 rebuild: 12-24 hours for a 3-4TB drive at typical rebuild speeds of 50-100 MB/s.
SATA SSD RAID 1 rebuild: 2-4 hours for a 1.92TB drive at 150-200 MB/s.
NVMe RAID 1 rebuild: Under 1 hour for a 3.84TB drive at sustained 1-2 GB/s rebuild speeds.
The rebuild speed matters because during rebuild, the surviving drive handles both production I/O and rebuild I/O simultaneously. The shorter the rebuild window, the less time the array spends in a degraded state where a second drive failure would cause data loss.
RAID Is Not Backup: The Critical Distinction
This distinction deserves explicit emphasis because the confusion is common and the consequences are severe.
RAID 1 protects against a single physical drive failure. It does not protect against:
Accidental file deletion (both drives delete the file simultaneously)
Database corruption from a software bug (both drives store the corrupted data)
Ransomware (both drives encrypt simultaneously)
Multiple simultaneous drive failures from a power surge or fire
Server theft or data center disaster
Protection against these failure modes requires backup to a separate physical location. InMotion Premier Care includes 500GB of automated backup storage stored off-server. This is the backup layer that complements RAID’s drive failure protection.
A complete data protection strategy uses both: RAID 1 for continuous drive fault tolerance with no downtime, and off-server backups for everything RAID cannot protect against. Neither replaces the other.