NUMA-Aware Thread Scheduling for Big Data Transfers over Terabits Network Infrastructure

Figure 2

Performance comparison for Memory-aware Thread Scheduling using Multiple Memory Buffers per CPU socket. The -axis shows the I/O thread count. N and NB are for baseline experiments. The error bar depicts min. and max. deviation from the average of 3 iterations.

(a) Big file workload

(b) Small file workload