How maxRAID Works: The Big Picture

How maxRAID Optimizes All-Flash Arrays

maxRAID is a host-based Flash Translation Layer (FTL) that optimizes all-flash arrays for performance and reliability. This brief explains how it works.

maxRAID’s Advanced Approach to WAF and Compression

At its core, maxRAID is a sophisticated block translation layer specifically designed to excel at managing garbage collection for random writes, a critical factor in reducing the Write Amplification Factor (WAF). WAF is an essential metric for the longevity of SSDs, as excessive write amplification can lead to premature wear and failure. As early as 2015, we employed a Monte Carlo simulator to model garbage collection and measure WAF under various conditions. The results of these simulations were validated by real-world maxRAID deployments, demonstrating that maxRAID consistently operates within 0.5% of the theoretical best-case WAF. This precise performance ensures optimal efficiency, helping organizations maximize the lifespan of their SSDs while minimizing write overhead.

Compression is another critical aspect of maxRAID’s design. By compressing data blocks, maxRAID significantly reduces write sizes, decreasing WAF and enhancing storage efficiency. This often allows WAF to drop below 1:1. Compression not only lessens write workloads but also creates additional free space, further lowering WAF. This dual benefit of reduced data size and increased free space directly translates into improved reliability, enhanced performance, and cost savings. As compression ratios improve, maxRAID dynamically adjusts to optimize storage capacity while maintaining peak performance, making it an ideal solution for evolving and unpredictable workloads.

Innovative Management of Empty Blocks

maxRAID also handles ‘empty’ blocks. While most SSDs offer basic trim/discard commands, maxRAID enhances this capability by automatically executing trim/discard operations whenever a block is written with all zeroes or ones. As a result, maxRAID works across various environments, including SANs and virtualized infrastructures, without the need for explicit support from clients or hypervisors. This universal compatibility removes potential integration hurdles and allows organizations to adopt maxRAID without revamping their existing storage architectures.

Optimized RAID Integration for Superior Performance

A key design choice in maxRAID is placing the FTL above the RAID layer, rather than within the SSD. When the FTL is placed below RAID, RAID operates as an ‘IO amplifier,’ effectively doubling or tripling the number of writes for RAID-10, RAID-5, or RAID-6 configurations. This also leads to additional reads for each write operation, further compounding the IO amplification issue. By relocating the FTL above the RAID layer, maxRAID ensures that RAID receives perfectly aligned, linear writes, significantly reducing IO amplification and enhancing overall system performance.

This advanced architecture improves write efficiency and enhances long-term SSD reliability by minimizing unnecessary write cycles and decreasing wear.

As Table 1 shows, this placement removes the read-modify-write penalty: maxRAID never has to read old data or parity before writing.

Table 1: I/O for each small random write

RAID level	Reads	Writes
Traditional RAID-5	2 (old data and parity)	2 (new data and parity)
Traditional RAID-6	3	3
maxRAID-5	none	1.15×
maxRAID-6	none	1.33×

Traditional figures are the standard read-modify-write cost of a small update. maxRAID figures are write amplification: for every 1 GB written, maxRAID-5 writes 1.15 GB to the SSDs and maxRAID-6 1.33 GB, as full sequential stripes.

Maximizing Media Lifespan and Efficiency with maxRAID

These benefits multiply rather than add up. Take data that compresses by 50%, on a fully utilized volume. For every 1 GB the application writes, maxRAID-5 writes about 0.75 GB to flash: 0.5 GB after compression, × 1.3 for garbage collection at 50% free space, × 1.15 for parity. RAID-10 writes 2 GB — every block is mirrored — before the SSDs’ own garbage collection adds more. Traditional RAID-5 also writes every small update twice, data and parity, and must read old data and parity first.

Media lifespan

Because maxRAID writes far less to each drive, SSDs last much longer than on standard RAID-5 — reducing replacement costs and improving the return on your storage investment. maxRAID adapts to different workloads and levels of data compressibility.

Future-Proofing Storage with maxRAID

maxRAID’s advantages extend far beyond wear reduction. By facilitating the use of parity RAID instead of mirroring, maxRAID significantly enhances storage capacity while ensuring data protection. Furthermore, its compression capabilities enable organizations to exchange a portion of wear endurance for increased usable space, providing a tailored solution to address specific application needs.

This intelligent design applies to all types of SSD media, including QLC, TLC, MLC, and even SLC. For instance, with far less wear than on RAID-5, QLC media may meet the write requirements of applications that once relied on more expensive TLC. Similarly, TLC can now act as a viable alternative where only SLC was previously considered sufficient, thereby reducing costs and enhancing storage flexibility. These options can lower media costs.

Conclusion: maxRAID’s Scalable and Reliable Storage Solution

These advantages are most apparent in workloads that are not solely read-heavy (featuring at least 0.3 writes per day) or exhibit some level of compressibility, especially in scenarios where writing is involved. For these workloads, maxRAID delivers more performance, capacity and drive life from the same SSDs.

To learn more, read the 200 TB Database Use Case.