Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Expose P99 latency and support customizable IO testing with selective IO types and metrics #9

Open
wants to merge 4 commits into
base: main
Choose a base branch
from

Conversation

derekbit
Copy link
Member

@derekbit derekbit commented Apr 7, 2024

  • Customizable IO testing with selective IO types and metrics
    Users have the option to select specific IO types, including random read (randread),
    random write (randwrite), sequential read (seqread), and sequential write (seqwrite),
    as well as metrics like IOPS, bandwidth, and latency.
    There's no requirement to conduct all tests.

    =========================
    FIO Benchmark Summary
    For: test_device
    CPU Idleness Profiling: disabled
    Size: 100M
    Quick Mode: disabled
    =========================
    IOPS in ops (Read/Write)
                      Random:           15,013 / 8,874
                  Sequential:          22,499 / 13,134
    
    Bandwidth in KiB/sec (Read/Write)
                      Random:        518,918 / 277,945
                  Sequential:        652,231 / 305,601
    
    Latency in ns (Read/Write)
                      Random:        298,378 / 419,719
                  Sequential:        295,757 / 414,594
    
    
    
    =========================
    FIO Benchmark Latency P99 Summary
    For: test_device
    CPU Idleness Profiling: disabled
    Size: 100M
    Quick Mode: disabled
    =========================
    Latency in ns (Read/Write)
                      Random:      667,648 / 1,028,096
                  Sequential:      618,496 / 1,003,520
    
  • Expose P99 latency
    Expose 99th percentile. It means that 99% of the requests should be faster than given latency.

longhorn/longhorn#8254

The CPU idleness introduces extra overhead and storage performance drop.
By default, the profiling is disabled.

Signed-off-by: Derek Su <[email protected]>
@derekbit derekbit marked this pull request as draft April 7, 2024 23:58
@derekbit derekbit marked this pull request as ready for review April 8, 2024 05:48
Users have the option to select specific IO types, including random read (randread),
random write (randwrite), sequential read (seqread), and sequential write (seqwrite),
as well as metrics like IOPS, bandwidth, and latency.
There's no requirement to conduct all tests.

Longhorn 8254

Signed-off-by: Derek Su <[email protected]>
Expose 99th percentile. It means that 99% of the requests should be faster than given latency.

Longhorn 8254

Signed-off-by: Derek Su <[email protected]>
Copy link
Contributor

@PhanLe1010 PhanLe1010 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks like this PR has big overlap with my other PR #7 . That other PRs also added many test modes.

Could we review and merge that PR first and build this one on top of it? @derekbit @shuo-wu @innobead. After the other PR is merged, I can update this PR to include the p99 if Derek is busy. Is it ok?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants