ESS w/ Compression / RAID

Traditional parity RAID suffers from "IO Amplification" for random updates. Whenever you write a block, the parity requires additional IO. This is additional writes, and additional reads.

Traditional parity RAID also has to deal with IO synchronization. When you update a block, you have to update the parity at the same time. This synchronization requirement is not possible without either leaving a "write hole", doing even more extra writes (a journal), or having power protected storage that requires specialized hardware.

ESS RAID avoids all of these issues by implementing an FTL in software on the host. This eliminates the extra IO of in-place random updates.

NVMe x8 Benchmarks. Run on AWS i4i.metal.

8 PCI-e gen-3 7500 GB NVMe SSDs. Tested with stock MD RAID-0/5/6. Tested with ESS Compressed RAID-0/5/6. Full details ... here ...

Compression lowers wear, increases performance, and gives you more usable space.

• 30% compressible data, get up to 40% extra storage.
• 60% compressible database, get up to 125% extra storage.
• You choose and tune extra space to fit your workload, wear, and performance needs.

Predicted SSD Life.

This is the predicted drive life for an 8-drive RAID-5 array. You can use the ESS RAID Designer Tool to caluclate array parameters with different drive counts and data compressibility.

ESS RAID lowers Flash wear in multiple, complimentary ways.

• RAID parity extra writes are eliminated.
• The ESS "FTL" has near "ideal write amp" based on free space.
• Block level compression increases free space.
• Block level compression further reduces wear.

Predicted SSD Life.

This is the predicted drive life for an 8-drive RAID-5 array. You can use the ESS RAID Designer Tool to caluclate array parameters with different drive counts and data compressibility.