White Papers

17
Figure 11. Sequential Write Performance
Figures 12 and 13 illustrate that F800 random read and write performance is better than Lustre and NSS for every test case. F800
random read performance is 7x greater than Lustre (140K vs 20K IOPS) at the 64-client case, while F800 random write performance is
3x greater than Lustre (45K vs 15K IOPS). Future tests will utilize more than 64 client compute nodes in an attempt to maximize F800
random read/write performance.