34 Replies to “New Radeon updates and what to do with them”

  1. Question … FreeBSD killer feature is zfs and arguably zfs is #1 reason why to use FreeBSD over another BSD.

    What’s dragonflybsd killer feature that separates it from all the other BSD?

    And don’t say Hammer, since a complete rewrite is underway to replace it with Hammer2 (with no eta on release)

  2. Well, OpenBSD “killer feature” is security and NetBSD “killer feature” is portability.

    I agree. Don’t know what DBSD is :(

  4. I don’t know if there is a single “killer feature” for DFBSD, but a lot of work (I think the initial impetus for the fork) has been done WRT scaling and performance. Easy enough to search the internet with your preferred search engine to get comparisons for networking and storage (like NvME) for comparison.

    Now is this the “defining” killer feature for DFBSD? No idea, I’m not speaking for the project, just as someone that’s been following a bunch of the BSDs for a long time.

  5. Why can’t I say Hammer? It’s here now.

    I remember some OpenBSD people saying “it’s not security, it’s correctness”.

  6. Justin

    If Hammer is so good then why is it being replaced by a radically different file system (hammer2)?

    Hammer doesn’t have CoW and many other ZFS-like features.

    I mean no disrespect when I say this … truly.

    But it just seems silly to me, like others, to say the primary feature as to why to use Dragonfly is also the same feature that for the last 6 years has had actively development to replace it

  8. Maybe the primary reason to use dragonfly specifically over FreeBSD, is “less politics”

    E.g. What happened to Dillon by FreeBSD team. DPorts. Etc.

  9. Anonymous – you are saying Hammer can’t be a primary feature to use because “Hammer 2” exists as a project? How does that make Hammer 1 filesystems not work?

    Who do you want to see updating those benchmarks? i.e. whose time are you volunteering, if not your own?

  10. Let me reframe what I think people are saying above … we have a product / market fit issue with DragonflyBSD.

    People don’t know *why* they should use DragonflyBSD over the other BSDs.

    With the other BSDs, you know why.

    OpenBSD = correctness & security

    NetBSD = portability

    FreeBSD = performance and ZFS

    Dragonfly = ???

    To be clear, I’m sure there’s dozens of reason why someone *should* use Dragonfly. We just need to collectively understand what those reasons are and market it

  11. DragonFly = Performance and Hammer (+modern graphic support)

    Hammer2 and Hammer are different monsters, and IMHO, they are targeting differently.

  12. Sepherosa

    What is Hammer and Hammer2 “targeting differently”?

    Will Hammer2 be as performant as Hammer?

    Do you view Dragonfly (currently) more performing that any other BSD for web server or database usage?

  14. @anonymous: funny, in the first benchmark from your Phoronix link, DragonFly is actually faster than all other operating systems, including the Linux-based ones.

  15. Anonymous

    In the perf benchmark above, 16 different speed tests were performed.

    Dragonfly was the fastest in only 1 of 16 test. And it just so happened to be the first test.

    In all of the other 15 test, dragonfly was not the fastest

  17. Why should the devs care?
    No one is going to pay them for marketing. Nor do I think a dev cares too much about marketing. The best thing about dbsd is that it (the sw) or the community suits you and your use case period.

  18. >> “The best thing about dbsd is that it (the sw) or the community suits you and your use case period.”

    You do realize that if people don’t even know WHY to use dbsd, that’s an issue.

    Educating people on why is called Content Marketing. And yes, educating people why to use your product is a form of marketing. Like it or not.

  20. Some funny reasoning here, I don’t want to use X because a newer better X+ is in the works. What happens when FreeBSD starts creating a “newer better” zfs. Everyone should bail on it?

  21. ZFS is not FreeBSD’s project. Also, ZFS project has died. ZFS project wanted to keep close to Solaris zfspool versionwise. FreeBSD was just closer to Solaris’s than other BSDs that also have ZFS as their feature. It’s just that FreeBSD wants to be good wrt marketing and politics. Do you see NetBSD advertizing the fact that zfs actually works on NetBSD x86 & amd64? Current FreeBSD’s ZFS is OpenZFS and is equally available to many, including to distros that are not using a bsd kernel, like linux. OpenZFS = zfspool version 5000. Every OpenZFS user has the same features! There are also the distros that use a kernel that is like linux, structurally, shared by several distros, each with their own userland, the illumos kernel. So, when will the nonsense stop, because ZFS is definitely not a unique feature of FreeBSD! DragonflyBSD is a lighterweight BSD that wants to be modern through simplicity. There are a ton of modern features that could get into DBSD but didn’t because of their overcomplexity and bloatness. ZFS won’t ever fit DragonflyBSD. DragonflyBSD inherited the performance oriented goals from FreeBSD, but nowadays it’s so easy to differentiate between these two performance BSDs. DragonflyBSD currently cares more about GCC, FreeBSD will stay with Clang. DragonflyBSD currently works with nginx reuseport, unlike FreeBSD. People who want modern as defined in FreeBSD won’t find DragonflyBSD interesting, so why advertize DragonflyBSD? Use it first if you think it’s for you. If you won’t ever have a need to install DragonflyBSD, why know it exists? It won’t fix your itches.

  22. I’m beginning to think we all pick the OS we want to use simply because of the logo/mascot.
    FreeBSD has Beastie (the new bobble thing sucks)
    DragonFly has a Red dragonfly
    OpenBSD has Puffy
    Linux has a penguin
    NetBSD has a flag
    Windows has a BSOD

  23. An active developer on board who cares about good available Ada support is enough for me to like dfly. That’s niche, but hey, every non-Windows OS is.

  24. For me, the killer feature is cheap filesystem-based deduplication.

    Being able to store 1.6TB of data on a 1TB volume hosted on a machine with only 4GB of memory is simply not possible with other operating systems.

  25. @Francois Tigeot

    Any chance of you updating your Postgres benchmarks?

  27. Francois benchmarks are the best. Really hope he updates his benchmarks as well (and includes Linux).

  30. The quickest, shortest path to a new set of benchmarks is to do them – and the methodology of the last set was documented.

  31. Maybe not all of have the proper skills to know how to perform a benchmark

  32. Pros and cons from one user’s point of view.

    Dragonfly’s biggest advantage is its friendly community and politics-free atmosphere. Dragonfly is also a very transparent project unlike a certain larger project that discusses items of major importance to the community in secret lists.

    Yes, Dragonfly is extremely performant, but you might not get all of the bells and whistles. One item that I’ve found lacking is monitoring capability, especially with regard to storage. If you’ve got a 30+ tb HAMMER volume, how on earth do you monitor it? HAMMER needs something like “zpool status” to get an overview of the integrity of the data. HAMMER “show” and the other relevant commands to do this take too long and don’t necessarily provide useful or actionable information.

    IMHO it is also somewhat disappointing is the fact that there is only one developer working on HAMMER2. Easy clustering is Dragonfly’s feature and storage clusters are becoming ever more important. Why isn’t H2 being actively worked on to get this very major feature done?

Comments are closed.