See that throughput is limited regardless of the backend storage capabilities.
Running 8 dd's simultaneously to the datastore.
However, even running only 1 dd, the throughput does not seem to go beyond a certain value.
No matter how many concurrent dd's are run, the throughput will not go beyond the high water mark set by running 1 dd copy process.
The dd tool is a built-in Linux tool that can be used to copy data to any block device like a disk or datastore.
VMware vSphere ESXi (All Versions)
The dd copies are not really simulating real workload as evidenced in the esxtop where the kavg and davg were nearly zero. While this can show if the datastore access works, it does not push enough data to the datastore to test for latency, throughput and IOPS.
Use HCIBench for a more real-world test of workload on a greenfield environment. Test with different block sizes and number of VMs and evaluate the performance after running the tests.
https://community.broadcom.com/vmware-cloud-foundation/discussion/command-line-tool-for-datastore-performance-test-in-esxi-6
https://community.broadcom.com/vmware-cloud-foundation/discussion/where-can-i-find-hcibench