[TASK] Add 1.5 performance benchmark to performance benchmark WIKI page

[innobead]

What's the task? Please describe

https://github.com/longhorn/longhorn/wiki/Performance-Benchmark

Need to verify v1 and v2, and do the comparison.

Describe the items of the task (DoD, definition of done) you'd like

Please use a task list for items on a separate line with a clickable checkbox https://docs.github.com/en/issues/tracking-your-work-with-issues/about-task-lists

• item 1

Additional context

cc @longhorn/dev-data-plane

[innobead]

cc @longhorn/qa

[derekbit]

We only measure the performance on high-level machines such as Equinix m3.small.x86.
As well as the high-level machines, we can benchmark on low- or medium-level machines like AWS c5.xlarge (4 vCPU, 8GiB RAM).

[shuo-wu]

Note: For the low-spec especially no dedicated disk machines, the performance gain is limited. Based on our testing result, the bandwidth of v2 volumes may be worse than v1 volumes:

One test result of a basic 8v CPU + 16G memory DigitalOcean machines:

================================
FIO Benchmark Comparsion Summary
For: Local-Path vs Longhorn
CPU Idleness Profiling: disabled
Size: 30G
Quick Mode: disabled
================================
                                   v1      vs                   v2        :              Change
IOPS (Read/Write)
        Random:            7,853 / 1,104   vs           10,920 / 3,339    :    39.06% / 202.45%
    Sequential:            9,371 / 2,293   vs           10,573 / 3,237    :     12.83% / 41.17%

Bandwidth in KiB/sec (Read/Write)
        Random:          53,113 / 13,652   vs          44,826 / 12,922    :    -15.60% / -5.35%
    Sequential:          43,148 / 13,903   vs          45,030 / 12,341    :     4.36% / -11.23%

Latency in ns (Read/Write)
        Random:    1,605,972 / 5,273,458   vs    1,079,956 / 1,308,515    :   -32.75% / -75.19%
    Sequential:    1,599,396 / 3,546,285   vs      738,651 / 1,227,585    :   -53.82% / -65.38%

[derekbit]

Note: For the low-spec especially no dedicated disk machines, the performance gain is limited. Based on our testing result, the bandwidth of v2 volumes may be worse than v1 volumes:

Similar results are observed on AWS platform. I've added a disk requirement in https://longhorn.io/docs/1.5.0/spdk/prerequisites/#disk.

[innobead]

@derekbit Let's update https://github.com/longhorn/longhorn/wiki/Performance-Benchmark first to include 1.5 performance v1 & v2.

[derekbit]

Updated. Thank you.