In Other BSDs for 2017/08/26

The new look on undeadly.org sure is nice.

9 Replies to “In Other BSDs for 2017/08/26”

  1. Re: BSD performance compared to highend router.

    The article states:

    “Finally, there is recent work in FreeBSD (which is part of 11.1-RELEASE) that gets performance up to 2x the level of netmap-fwd or the work by Nanako Momiyama. Here is a decent introduction.”

    Given the recent network perf benchmark that Sepherosa Ziehau completed and it showing FreeBSD so poor

    https://leaf.dragonflybsd.org/~sephe/perf_cmp.pdf

    https://www.dragonflydigest.com/2017/03/06/19425.html

    I wonder how much of an improve FreeBSD networking is now

  2. Re: BSD performance compared to highend router.

    The article states:

    “Finally, there is recent work in FreeBSD (which is part of 11.1-RELEASE) that gets performance up to 2x the level of netmap-fwd or the work by Nanako Momiyama. Here is a decent introduction.”

    Given the recent network perf benchmark that Sepherosa Ziehau completed and it showing FreeBSD so poor

    https://www.dragonflydigest.com/2017/03/06/19425.html

    I wonder how much of an improve FreeBSD networking is now

  3. I recently completed the configurable # of netisrs, which is intended to improve IP forwarding performance if there is enough NIC RX rings. Don’t have time to do the complete measure yet.

  4. Dfly has gotten so strong lately, I have to wonder … is there any longer a major reason why someone should use FreeBSD over Dly anymore (for web server and or database OS)?

    Dly truly seems to do it all, and typically does it better than FreeBSD.

  5. Dflybsd seems to be equally strong and often stronger in a lot of areas

  6. I would like someone with a specific work scenario to test it with different BSDs, and publish it. That’s a LOT of labor to produce, though.

  7. What’s fantastic about the slides from 2014 below that Francois Tigeot completed is it perf benchmarking database (Postgres) load on various BSD and Linux

    http://www.pgcon.org/2014/schedule/attachments/309_PG_as_bench_mark.pdf

    In my opinion these type of common-use-case benchmarks are super important not because we’re trying to see who’s OS is a few percentage points faster but more so to:

    1. To ensure the OS isn’t regressing
    2. To ensure the OS is performing relative to other OS (e.g. if one OS is way below other OS in perf, that should be investigated)
    3. To provide a relative gauge on the merits of how one OS implements a feature and what perf impact that has in comparison to other OS

    I really hope Francois can find time to update these slides.

Comments are closed.