I have a samsung T5 SSD hooked up to the USB port of a rb5009ug+s+in. On an otherwise idle network with a 10gb/s connection between a windows machine and the rb5009 I see 1.3 gb/s (160 MB/s) both for reads and writes over smb. This particular drive can achieve 4 gb/s read and 2.5+ gb/s write when hosted on a windows machine over smb. I looked at the rb5009 utilization and see 1-2 cpus for both read and write and irqs for writes near 100%.
A speed test from the rb5009 to the windows machine achieves the following. I know it's not ideal to run speedtest on the rb5009 since that will add overhead which limits throughput but I don't currently have another option. This is mainly to show the link and intermediate switches can achieve higher throughput.
TCP Download 3.93Gbps local-cpu-load:97%
TCP Upload 4.22Gbps local-cpu-load:98% remote-cpu-load:0%
I know it's unrealistic to expect the rb5009 to saturate the usb3 port or drive over smb since it's not designed to be a nas. However, given that all cpu cores are not used and assuming that's limiting performance, can throughput be improved with router os updates?
Download from rb5009
Upload to rb5009