In theory, an array that consists of 16 flash SSDs capable of delivering 200 MB/s each should be able to get to 3,200 MB/s. Unfortunately, we remained far away from this number despite all optimizations, but we still managed to outperform Samsung’s flash SSD array. That array was powered by 24 PB22-J flash SSDs. We only used 16 Intel X25-E SSDs to reach higher numbers. The primary goal was to achieve a higher sustained sequential throughput, and we’re very pleased to have hit 2.23 GB/s with our X25-E RAID versus 2.12 GB/s for the Samsung PB22-J array.
Bottlenecks can most likely be found in CPU performance as well as farther down the platform in the storage controllers. Although the theoretical bandwidth of each PCI Express x8 interface should be 2 GB/s, real throughput numbers might be much lower.
Should you have feedback or recommendations as to how you’d try to improve performance, please let us know and comment on this article!