RAID Rebuild Guide for the HP ProLiant DL380 Gen9

Rebuild Times, IOPS Expectations, and RAID Controller Best Practices

For many IT professionals, small businesses, and home lab enthusiasts across the UK, the HP ProLiant DL380 Gen9 remains one of the most trusted enterprise servers available. Even years after its original release, it continues to be widely used in laboratories, small server rooms, and edge deployments thanks to its impressive reliability, flexible configuration options, and excellent performance when purchased on the refurbished market.

One of the most common storage configurations for this server involves 900GB 2.5-inch 10K SAS drives. These drives offer a strong balance between performance and storage capacity while maintaining the durability expected of enterprise hardware. For organisations running virtual machines, shared storage systems, or backup infrastructure, such drives provide dependable performance without the cost associated with large all-flash arrays.

However, building a reliable RAID storage system is about much more than simply installing drives and selecting a RAID level. Real-world storage performance depends on several technical factors, including rebuild behaviour, controller settings, workload patterns, and disk performance characteristics.

Many administrators only begin to think about these issues after something goes wrong, such as when a drive fails and the array enters a degraded state. Understanding how RAID behaves before problems occur allows you to design storage systems that are both faster and more resilient.

In this in-depth guide we will explore several topics that are often overlooked when designing RAID storage:

• RAID rebuild times and why they matter
• Disk IOPS expectations in various RAID levels
• RAID controller configuration best practices
• Mechanical SAS versus SSD performance
• Real-world storage layouts for different workloads
• Monitoring and maintenance strategies for long-term reliability

To make the discussion practical, we will refer to two common storage configurations frequently seen in the HP ProLiant DL380 Gen9 platform.

Example Storage Configurations

Configuration A

• 8 × 900GB SAS 10K drives
• 7.2TB total raw capacity

Configuration B

• 24 × 900GB SAS 10K drives
• 21.6TB total raw capacity

Both systems utilise the integrated enterprise RAID controller typically shipped with this platform: the HPE Smart Array P440ar Controller.

This controller supports a wide range of RAID levels, caching options, and performance optimisations that make it suitable for demanding enterprise environments.

Before diving into performance expectations, let us begin with one of the most important — and frequently misunderstood — aspects of RAID storage: rebuild times.

---

Understanding RAID Rebuild Times

Even the most carefully designed storage infrastructure cannot avoid hardware failures entirely. Mechanical disks contain motors, bearings, and read/write heads that eventually wear out. While enterprise SAS drives are designed for high reliability, failures still occur over time.

RAID technology protects data by spreading information across multiple disks while maintaining redundancy. If a disk fails, the RAID controller can reconstruct the missing data using information stored on the remaining drives.

This reconstruction process is called a RAID rebuild.

A rebuild typically begins when:

• A failed disk is replaced by an administrator
• A hot spare drive automatically takes over
• A previously offline disk returns to the array

During this process, the RAID controller reads data blocks from the surviving disks and reconstructs the missing information onto the replacement drive.

Although this process is automatic, it can take a considerable amount of time.

---

Factors That Influence Rebuild Duration

Rebuild times vary depending on several technical factors:

1. Disk Capacity

Larger disks naturally take longer to rebuild because more data must be reconstructed. Modern enterprise drives can exceed 10TB, making rebuild operations significantly longer than in earlier generations.

2. Disk Speed

Disk rotational speed and interface bandwidth influence how quickly data can be read and written during the rebuild.

10K SAS drives typically offer much higher throughput than traditional 7.2K SATA disks.

3. RAID Level

Different RAID levels reconstruct data in different ways. Parity-based RAID levels require additional calculations during rebuild operations.

4. System Workload

If the server is actively serving applications or virtual machines during a rebuild, the controller must divide its resources between normal operations and the reconstruction process.

5. RAID Controller Performance

The processing power and cache capacity of the RAID controller also play a role. Enterprise controllers such as the HPE Smart Array P440ar are designed to handle rebuild operations efficiently.

---

Throughput of 10K SAS Drives

A typical 900GB 10K SAS drive in good condition can sustain approximately:

140–180 MB/s sequential throughput

However, rebuild operations rarely achieve this maximum throughput because:

• Multiple disks must be read simultaneously
• Parity calculations consume controller resources
• Workload traffic competes for bandwidth

As a result, rebuild speeds are usually slower than raw disk throughput would suggest.

---

Typical Rebuild Times for 900GB Drives

The following table provides approximate rebuild durations under relatively light system load:

These estimates assume the server is not under heavy utilisation.

In real production environments, rebuild times can increase significantly.

For example, rebuilding a failed disk in a 24-disk RAID 6 array within an HP ProLiant DL380 Gen9 could easily take eight hours or longer, especially if the server is actively running virtual machines or backup jobs.

---

The Hidden Risk During RAID Rebuilds

One of the most important aspects of rebuild operations is the risk window that occurs while the array is degraded.

When a disk fails, the RAID array temporarily loses part of its redundancy. Until the rebuild completes, the system becomes more vulnerable to additional failures.

Consider the following examples.

RAID 5

RAID 5 can tolerate one disk failure. If another disk fails during the rebuild, the array is likely to suffer catastrophic data loss.

RAID 6

RAID 6 can tolerate two disk failures, which significantly reduces the risk during rebuild operations.

RAID 10

RAID 10 mirrors disks in pairs, so the array can survive multiple failures as long as both disks in a mirror pair do not fail simultaneously.

Because modern arrays often contain dozens of disks, the probability of a second failure during rebuild becomes increasingly important.

This is one of the reasons why RAID 5 is gradually disappearing from enterprise storage deployments in favour of RAID 6 or RAID 10.

---

Understanding Disk IOPS Performance

While rebuild times determine how quickly arrays recover from failure, IOPS performance determines how fast applications can access data.

IOPS stands for Input/Output Operations Per Second, and it measures the number of read or write operations that a storage device can perform each second.

This metric is particularly important for workloads involving many small random operations, such as:

• Virtual machine storage
• Database transactions
• Email systems
• Web application servers

Mechanical disks are inherently limited by physical motion.

Each operation requires the disk head to move to a specific location on the platter, introducing latency.

---

Typical IOPS of 10K SAS Drives

A standard enterprise 10K SAS disk usually delivers:

120–180 random IOPS

For estimation purposes, administrators often assume an average of 150 IOPS per disk.

However, RAID configuration significantly influences the total usable IOPS of an array.

---

RAID Write Penalties Explained

Parity-based RAID levels introduce additional overhead for write operations.

When data is written to a RAID 5 or RAID 6 array, the controller must update both the data and the parity information that protects it.

This process involves several steps:

1. Reading the existing data block
2. Reading the existing parity block
3. Calculating new parity
4. Writing the updated data block
5. Writing the updated parity block

Because of these extra operations, each write request generates multiple disk operations.

Typical write penalties include:

• RAID 5 → 4 I/O operations per write
• RAID 6 → 6 I/O operations per write

This significantly reduces effective write performance.

---

Estimated IOPS for an 8-Disk Array

Using the earlier assumption of 150 IOPS per disk, we can estimate performance for an eight-disk configuration.

RAID 10 offers strong performance because it combines striping and mirroring without requiring parity calculations.

For workloads with heavy write activity, RAID 10 often performs significantly better than RAID 5 or RAID 6.

---

Performance in Large Disk Arrays

Larger arrays naturally increase IOPS capacity.

For a 24-disk configuration, approximate performance might look like this:

RAID 50 and RAID 60 divide disks into smaller parity groups, which improves performance compared with a single large RAID 5 or RAID 6 set.

Nevertheless, RAID 10 still offers the strongest write performance in most situations.

---

RAID Controller Configuration Best Practices

Even the best RAID layout can perform poorly if the controller is not configured correctly.

The HPE Smart Array P440ar Controller offers several advanced settings that can significantly affect performance.

Understanding these settings is essential when deploying enterprise RAID storage.

---

Write Cache Configuration

One of the most important features of enterprise RAID controllers is write-back caching.

When enabled, write operations are temporarily stored in the controller’s cache memory before being written to disk. This allows the controller to combine multiple small operations into larger sequential writes.

Advantages include:

• dramatically improved write performance
• reduced disk latency
• smoother performance during heavy workloads

However, write-back cache should only be enabled when the controller has battery-backed or capacitor-backed cache protection.

This ensures that cached data is not lost during a power failure.

Without this protection, the controller will typically revert to write-through mode, which writes data directly to disk and is significantly slower.

---

Read Cache and Adaptive Read Ahead

Read caching can also improve performance, particularly for workloads involving sequential data access.

Adaptive read ahead allows the controller to detect sequential read patterns and automatically preload additional data blocks into cache.

This is especially beneficial for:

• backup operations
• file transfers
• media streaming
• large archive reads

For most workloads, adaptive read ahead is a safe and effective configuration.

---

Stripe Size Considerations

Stripe size determines how data blocks are distributed across disks within a RAID array.

Choosing the correct stripe size can improve performance depending on the workload.

Typical recommendations include:

Smaller stripes improve random access performance, while larger stripes favour sequential workloads.

For many home labs and small business environments, 64KB or 128KB provides a good balance.

---

SSD vs SAS RAID Performance

Over the past decade, solid-state drives have dramatically changed the storage landscape.

Many home labs now upgrade the HP ProLiant DL380 Gen9 with SSD storage instead of mechanical disks.

The performance difference is substantial.

A typical SATA SSD can deliver:

50,000–90,000 IOPS per drive

In comparison, a mechanical 10K SAS disk provides only around 150 IOPS.

This means that even a small SSD array can easily outperform a large SAS array.

Example comparison:

However, mechanical disks still remain relevant.

They offer:

• lower cost per gigabyte
• predictable sustained throughput
• excellent endurance for archival workloads

Because of these advantages, many administrators use hybrid storage architectures.

---

Hybrid Storage Strategies

A common enterprise design combines the strengths of both technologies.

For example:

• SSDs store active virtual machines or databases
• SAS drives store bulk data, backups, and archives

This approach balances performance and cost while providing large storage capacity.

---

Recommended RAID Layouts for Different Workloads

Different workloads have different storage requirements.

Selecting the correct RAID layout can greatly improve system behaviour.

---

Virtualisation Servers

Virtual machine platforms such as VMware ESXi generate large numbers of small random disk operations.

A common layout is:

• RAID 1 for operating system boot drives
• RAID 10 for VM datastores

Advantages include:

• excellent random I/O performance
• rapid rebuild times
• strong redundancy

---

File Storage Systems

File storage platforms such as TrueNAS often prioritise capacity and reliability.

For these systems, RAID 6 is frequently recommended.

Benefits include:

• protection against two disk failures
• efficient use of disk capacity
• good sequential read performance

---

Backup Repositories

Backup platforms such as Veeam Backup & Replication typically require large storage capacity and strong fault tolerance.

Recommended RAID configurations include:

• RAID 6
• RAID 60

These provide strong protection against disk failures while maintaining reasonable performance for sequential workloads.

---

Monitoring and Maintenance

Even the most carefully designed RAID array requires ongoing monitoring.

Administrators should implement the following best practices:

• Enable RAID controller alerts
• Replace failed disks immediately
• Maintain at least one spare drive
• Perform regular consistency checks
• Monitor disk SMART statistics

The HP ProLiant DL380 Gen9 includes a powerful remote management interface known as HPE Integrated Lights-Out, often called iLO.

This interface allows administrators to:

• monitor hardware health
• receive alerts for disk failures
• manage the server remotely
• access the system console

iLO greatly simplifies server management, particularly in remote deployments.

---

Final Thoughts

Designing an effective RAID storage system requires careful consideration of several factors:

• performance requirements
• disk capacity
• rebuild times
• redundancy levels
• workload characteristics

The HP ProLiant DL380 Gen9 remains a capable and flexible platform that can support a wide range of RAID configurations.

For most small businesses and home labs, the following general guidelines work well:

• RAID 10 for performance-critical workloads
• RAID 6 for large storage arrays
• RAID 50 or RAID 60 for very large disk groups

It is also essential to remember that RAID is not a backup solution.

RAID protects against hardware failure but does not safeguard against accidental deletion, ransomware, or software corruption.

A complete data protection strategy should include:

• regular backups
• off-site storage
• proactive monitoring
• routine maintenance

With proper planning and configuration, RAID arrays can provide many years of reliable service. Combined with robust enterprise hardware such as the HP ProLiant DL380 Gen9, they remain a powerful foundation for modern storage infrastructure.