maxRAID is a host-based Flash Translation Layer (FTL) that optimizes All-Flash Arrays. But how does it deliver such impressive benefits? It all comes down to intelligent design and math.
At its core, maxRAID is a block translation layer that excels in handling garbage collection for random writes, reducing the Write Amplification Factor (WAF)—a key metric for SSD longevity. As early as 2015, we used a Monte Carlo simulator to model garbage collection and WAF. The results were validated by real-world maxRAID deployments, showing that maxRAID operates within 0.5% of the theoretical best-case WAF. For any given amount of free space, simply subtract 0.5% from the theoretical best WAF, and you have the maxRAID WAF.
Compression is another significant factor. By compressing blocks, maxRAID reduces the size of writes, further decreasing WAF. This often drives WAF below 1:1. Compression creates more free space, lowering WAF and improving efficiency.
maxRAID also intelligently handles “empty” blocks. While most SSDs support trim/discard commands, maxRAID takes it further by automatically applying trim/discard whenever a block is written with all zeroes or ones. This works universally across SANs and virtualized environments without explicit support from clients or hypervisors.
Finally, maxRAID optimizes RAID by positioning the FTL above the RAID layer rather than inside the SSD. When the FTL is below the RAID, RAID acts as an “IO amplifier”—doubling or tripling writes for RAID-10, RAID-5, or RAID-6 and adding extra reads for each write. Moving the FTL above RAID, maxRAID ensures that RAID sees perfectly aligned linear writes, drastically reducing IO amplification, as shown in Table 1.
These benefits multiply rather than simply add up. For 50% compressible data compared to RAID-10 on a full volume:
Standard RAID-5 has a WAF six times worse than RAID-10—calculated as 3:1 for 1 DWPD, doubled due to mirroring.
maxRAID-5, with 50% compressible data, achieves a WAF 0.75 times better than RAID-10—calculated as 0.5:1, multiplied by 1.3 for maxRAID’s WAF at 50% free space, multiplied by 1.15 for RAID-5’s IO overhead.
In addition, maxRAID reduces wear by 8x, more than 9 years compared to RAID-5. While these calculations are simplified, maxRAID seamlessly adapts to different workloads and varying data compressibility.
The benefits extend beyond wear reduction. By using parity RAID instead of mirroring, maxRAID boosts capacity. Compression also allows you to trade some wear endurance for additional usable space based on your application’s needs.
This math applies to any SSD media—QLC, TLC, MLC, or even SLC. For example, QLC at 8x improvement over RAID-5 may meet your write requirements, allowing you to replace TLC in some applications with less expensive alternatives. Similarly, TLC might now be suitable where SLC was previously needed.
These advantages shine in workloads that aren’t read-heavy (at least 0.3 writes per day) or have some compressibility, especially when write performance is critical.
