I was running some performance benchmarks using fio to compare the performance of a Docker volume (created using the -v option) against the performance of a volume mounted on the host machine. I am using the following command to run IO on both the volumes:
fio --name=seqwrite --ioengine=libaio --iodepth=1 --rw=write --bs=8M --direct=1 --size=1024M --numjobs=24 --runtime=300 --group_reporting
For the host-mounted volume I am using a single SATA drive (2TB) formatted using ext4.
The throughput of the Docker volume is around 60MB/s as compared to the host-mounted volume which is 125-130 MB/s. Also, when I bind mount the same host volume inside the Docker container, I get the same 125-130 MB/s throughput. Is there a known performance degradation issue when we create Docker volumes using the "-v" option or is there anything incorrect with my methodology of testing and comparing the performance?
