We have been using the kbench tool to benchmark topolvm against longhorn with data locality set to best-effort.

================================
FIO Benchmark Comparsion Summary
For: Topolvm vs Longhorn
CPU Idleness Profiling: disabled
Size: 30G
Quick Mode: disabled
================================
                              Topolvm   vs                 Longhorn    :              Change
IOPS (Read/Write)
        Random:         121,199 / 97,226   vs          25,142 / 13,671    :   -79.26% / -85.94%
    Sequential:         118,515 / 98,933   vs          40,541 / 13,730    :   -65.79% / -86.12%

Bandwidth in KiB/sec (Read/Write)
        Random:    4,873,592 / 3,993,511   vs         166,344 / 55,488    :   -96.59% / -98.61%
    Sequential:    6,013,546 / 4,016,189   vs         158,778 / 55,382    :   -97.36% / -98.62%

Latency in ns (Read/Write)
        Random:          80,875 / 21,071   vs        570,239 / 857,724    :  605.09% / 3970.64%
    Sequential:          18,339 / 20,979   vs        576,889 / 836,005    : 3045.69% / 3884.96%

it’s really not bad but I was expecting longhorn to perform a bit better than it did. Maybe there are some things I can tweak?

v2 engine is faster, but not prod-ready

I tried to use it but didn't work for us, kept the instance manager kept crashing