Solid state drives (SSDs) are becoming increasingly popular for use in storage arrays and servers due to their performance advantages over traditional hard disk drives (HDDs). SSDs offer much faster read/write speeds, lower latency, and reduced power consumption. However, there are some unique considerations when using SSDs in a RAID configuration. In this article, we’ll examine the pros and cons of using SSDs for RAID and discuss best practices.
The Benefits of SSDs for RAID
There are several potential benefits to using SSDs in a RAID array:
- Increased performance – SSDs provide much faster access to data than HDDs. This can result in dramatically improved throughput for I/O intensive applications.
- Lower latency – The nature of NAND flash used in SSDs means they can access data almost instantaneously. Average latency is in the microsecond range versus milliseconds for HDDs.
- Better IOPS – SSDs can achieve tens of thousands to hundreds of thousands of input/output operations per second (IOPS). HDDs are typically limited to hundreds of IOPS.
- Reduced power consumption – SSDs use less power than HDDs both active and idle. This can help reduce data center operating costs.
- Increased resilience – SSDs have no moving parts and can better withstand physical shocks. They may be less likely to fail in arrays that experience vibration in racks.
- Compact size – 2.5″ SSDs take up less space which can be useful in high density servers and arrays.
The performance advantages of SSDs become even more pronounced when arranged in a RAID configuration. The parallel nature of spreading I/O across multiple drives plays well to the strengths of SSDs. Ultimately, SSDs can provide substantially higher performance levels in RAID compared to HDDs.
Considerations When Using SSDs for RAID
However, there are some unique considerations to factor in when using SSDs for RAID:
- Cost – SSDs still carry a price premium over HDDs. Populating an entire array with SSDs will be very expensive.
- Lifespan – SSDs can only write a finite number of times to each memory cell before wearing out. RAID configurations may wear them out faster.
- TRIM support – TRIM allows the SSD to efficiently handle garbage collection. Some RAID controllers do not yet support passing along TRIM commands.
- RAID rebuild times – Rebuilding a failed drive means writing the full capacity which can wear out SSDs more quickly.
- RAID write penalty – Many RAID levels incur a write penalty which magnifies the limited endurance of SSDs.
These factors require special consideration to ensure proper SSD lifespans and performance in a RAID implementation.
SSD RAID Performance
To achieve maximum performance from SSD RAID, it is important to select SSDs designed for enterprise workloads. Consumer SSDs often do not have capacitors to flush cached data to NAND on sudden power loss. Enterprise SSDs support interfaces like SAS and NVMe which offer better performance than SATA. Optimal SSD RAID performance also requires using an appropriate RAID controller that supports the SSD’s interface and relevant features like TRIM.
SSD RAID performance can scale nearly linearly with each drive added. Testing by StorageReview showed random 4K read performance scaling from 245,000 IOPS with one Intel SSD to 965,000 IOPS in a RAID 10 array of eight drives.1 But performance characteristics can vary based on the RAID level used:
- RAID 0 – Block-level striping provides the best overall performance but no redundancy. Latency is excellent thanks to parallel spreads of I/O.
- RAID 1 – Disk mirroring provides redundancy at the expense of usable capacity. Read performance can scale but writes do not improve.
- RAID 5 – Block-level striping with distributed parity has middling performance but can recover from a single drive failure. Writes slow down due to parity calculation overhead.
- RAID 6 – Similar to RAID 5 but can withstand the loss of two drives. Write performance takes an even bigger hit due to more parity calculations.
- RAID 10 – Striping and mirroring together provide the best balance of speed and redundancy. Allows sequential read/writes to scale linearly while tolerating multiple drive failures.
In most cases, RAID 10 provides the best blend of high performance and fault tolerance for mission critical SSD RAID implementations.
SSD RAID Setup Considerations
Properly configuring SSD RAID requires paying attention to the following factors:
- RAID Controller – Select an enterprise-class hardware RAID controller that supports the SSD’s interface and has at least 1GB of DRAM cache. Make sure the firmware is up to date with SSD support.
- Drive Interface – SAS and NVMe SSDs provide better performance than SATA in RAID scenarios. NVMe is emerging as the top choice thanks to phenomenal throughput from PCIe connectivity.
- TRIM Support – Ensure TRIM commands can be passed from the operating system through the RAID controller to the SSDs.
- RAID Level – RAID 10 provides the best overall blend of performance and drive failure protection for SSD RAID.
- Spares – Keep ample spare SSDs on hand for rapid replacement in case of drive failure.
- Monitoring – Track SSD lifespan metrics like wear levels and uncorrectable errors to identify potential drive failures before they happen.
Paying attention to these aspects will help maximize the lifetime and performance of an SSD RAID deployment.
Is SSD RAID Worth It?
SSD RAID can deliver substantial performance advantages over HDD RAID in terms of throughput, latency and IOPS. But is the additional cost and complexity worth it? Here are a few key factors to consider:
- For transactional databases and other highly I/O intensive applications, the speed boost of SSD RAID may justify the premium expense.
- If application performance is limited by disk I/O, SSD RAID can remove storage bottlenecks to improve overall efficiency.
- For mission critical systems requiring maximum uptime, the redundancy of RAID 10 with SSDs can provide necessary data protection.
- If power consumption, heating and data center footprint are concerns, the superior density of SSD RAID can help.
- Consumer SSDs in RAID have more drawbacks around endurance and reliability. Enterprise SSDs are preferred for RAID use cases.
- For less demanding applications like file servers, HDD RAID may still offer the best balance of cost and performance.
In the right scenarios, the benefits of SSD RAID can make it well worth the investment. As SSD costs continue to decrease over time while capacities increase, adopting SSD RAID will make economic sense for a broader range of workloads.
Best Practices for SSD RAID
To get the most out of an SSD RAID deployment, keep these best practices in mind:
- Use enterprise SSDs designed for mixed workloads rather than consumer SSDs.
- Select SSDs in a consistent capacity and form factor to simplify RAID management.
- Choose an enterprise RAID controller that supports SSD features like TRIM.
- Enable TRIM support in the operating system, RAID controller firmware and SSD firmware.
- Use higher RAID levels like RAID 10 for mission critical storage to balance performance and drive failure tolerance.
- Monitor SSD wear indicators and replace drives proactively before failures occur.
- Spread read/write demand evenly across drives to prevent uneven wear levels.
- Ensure proper cooling and airflow for SSDs which run hotter than HDDs under load.
- Keep firmware up to date on SSDs and the RAID controller.
With the right RAID implementation following these guidelines, SSD RAID can deliver exceptional performance for I/O intensive workloads.
RAID Controller Considerations for SSDs
The RAID controller plays a critical role in extracting the best performance from SSDs in a RAID configuration. Here are some key factors to evaluate when selecting a RAID controller for SSD RAID:
- Interface support – Choose a RAID card that supports the fastest drive interfaces used by the SSDs, such as SAS, SATA, or NVMe.
- SSD-aware features – Seek out RAID cards with SSD optimization features like TRIM passthrough and write serialization.
- Cache memory – Larger DRAM caches of 1GB or more help absorb write strain and improve read performance.
- Bandwidth – 12Gbps SAS and PCIe 3.0 x8 bandwidth or higher offer fast connectivity for peak SSD throughput.
- RAID algorithms – Cards optimized for SSDs use less redundant parity calculations to reduce writes.
- Power protection – Sudden power loss can corrupt SSD data, so look for capacitors and flash backups.
Top examples of RAID cards purpose-built for optimizing SSD RAID performance include offerings from Broadcom, LSI, Microsemi, Intel, and Dell. When shopping for a RAID controller, be sure to verify that SSD-specific capabilities are supported before purchase.
SSD RAID Reliability and Redundancy
Data protection is a top priority for any storage system. SSD RAID can provide improved reliability and redundancy versus a standalone SSD, but potential drawbacks around drive lifespan need to be factored in.
With HDD RAID, the likelihood of simultaneous failures across multiple drives is very low. But SSDs have a finite lifespan limited by program/erase cycling. When using RAID to span SSDs, the chances of correlated failures rises as multiple drives wear out from heavy use. To counter this, SSD RAID needs even more rigorous monitoring and proactive replacement compared to HDD RAID.
SSD RAID can also exacerbate the wear on drives compared to standalone SSDs. The constant parity writes in RAID 5/6 and mirror writes in RAID 1/10 cause substantial write amplification. Consumer-grade SSDs will fare worse in this environment while enterprise SSDs with higher endurance ratings are preferred.
To account for these factors, a wise approach is to use enterprise SSDs rated for 10 or more drive writes per day (DWPD) and opt for higher RAID levels like RAID 10. Excellent monitoring tools are key to track SSD wear and lifecycle metrics. Following these reliability best practices allows SSD RAID to provide excellent redundancy for critical data.
Comparing HDDs vs. SSDs for RAID
| Hard Disk Drives (HDD) | Solid State Drives (SSD) | |
|---|---|---|
| Price per GB | Much lower | Higher |
| Lifespan | Typically 5+ years | Finite write endurance |
| Performance | Lower throughput Higher latency |
Faster throughput Lower latency |
| Power Usage | Higher active power Lower idle power |
Lower active power Higher idle power |
| Failure Tolerance | High MTBF Lower UBER |
Potentially higher UBER |
| Noise | Audible head movements | Silent operation |
| Shock Resistance | Contains moving parts | No internal moving parts |
While HDD RAID maintains advantages in capacity and affordability, SSD RAID offers superior performance, resilience, power efficiency and footprint. In data centers where speed, uptime and density matter most, SSDs are increasingly becoming the preferred RAID storage medium.
SSD RAID Configuration Scenarios
SSD RAID can enable high performance storage in a variety of usage scenarios. Some common applications that lend themselves well to SSD RAID include:
- Online Transaction Processing (OLTP) Databases – Databases supporting high volumes of transactions require very fast I/O. SSD RAID delivers consistent low latency and high IOPS to significantly improve query performance.
- High Performance Compute (HPC) – HPC workloads like genomic sequencing, machine learning and financial modeling depend on massive parallel I/O. SSD RAID can feed data to compute nodes at speeds far exceeding HDD RAID.
- Virtual Desktop Infrastructure (VDI) – Providing fast desktop experiences to hundreds of VDI users requires high IOPS and low latency from the shared storage. SSD RAID meets demands even under heavy load.
- High Frequency Trading – Shaving off milliseconds of latency for transactions can mean big profits. The blistering speed of SSD RAID delivers the edge financial trading firms need.
- Media Post-Production – Video editing teams require storage that can playback raw 4K/8K footage smoothly. SSD RAID provides the throughput to edit high-res projects fluidly.
In each case above, HDD RAID would likely struggle to meet the required performance levels due to disk speed limitations. The ultra-fast access of SSD RAID makes it shine for these demanding workloads.
Key Takeaways
- SSD RAID can deliver substantial performance gains for I/O intensive applications, but at increased cost over HDD RAID.
- Enterprise SSDs with higher endurance are strongly preferred for use in RAID over consumer SSDs.
- RAID levels like RAID 10 provide an ideal blend of performance and fault tolerance for SSD RAID.
- Proper RAID controller selection helps maximize the speed and lifespan of SSD RAID.
- Compared to HDD RAID, SSD RAID offers significantly faster throughput and lower latency but requires extra care around drive lifespans.
- For transactional databases, virtualized environments, HPC and other performance-sensitive applications, SSD RAID can be compelling versus large HDD RAID.
While SSD RAID has higher upfront costs, the dramatic speed benefits for I/O-intensive applications make it an investment well worth considering.
Conclusion
SSD RAID opens new horizons for storage performance that is simply not possible with HDD RAID. By combining the parallelism of RAID with the lightning fast access of solid state drives, SSD RAID can keep pace with the growing demands of today’s most demanding workloads. With proper tuning and setup, organizations can benefit from SSD RAID’s winning combination of competitive economics, unmatched speed, and robust data protection.
