Skip to main content
. 2023 Jul 27;39(8):btad463. doi: 10.1093/bioinformatics/btad463

Figure 2.

Figure 2.

Benchmarking results of various operating system IO APIs. Sequential reads of 4k blocks were carried out using POSIX, memory map (mmap), and libaio on both NVME SSD available on AWS and a Lustre distributed file system available to our local compute cluster. We in addition benchmarked io_uring but only on AWS as it is supported by only recent Linux kernels not yet available at our local facility.