The question of using vkernels(7) in a manner similar to jails pops up time and again, and the answer is, unsurprisingly, “it depends“. It looks like when you want to isolate greedy programs, vkernels are the way to go.
38 Replies to “Vkernels: when to use them”
Comments are closed.
Justin, do you have any insight in what the next release of DBSD will be (e.g. 4.8 or 5.0) and also what major features will it include?
+1 updated roadmap.
The current roadmap is way out of date.
https://bugs.dragonflybsd.org/projects/dragonfly/roadmap
And I recall reading a post to the list serv last year entitled “DragonFlyBSD goals and roadmap for next decade” in which no one responded.
My guess would be 4.8 is the next release. Releases are roughly every 6 months, and we’re only 3-4 months into this one, so it will be a while.
The work Matthew Dillon has been doing on the kernel is pretty major, and I haven’t been covering it as thoroughly as I’d like – but it’ll be there in the next release.
Hank – DragonFly is a loose group of volunteers working on one of the last truly noncommercial operating systems. The group isn’t really structured the right way for a roadmap to work, without someone with the natural authority to push it. Which is okay!
Well said Justin, although this “non-commercial” operating system will have tremendous commercial value in when the HAMMER2 pieces are ready. In fact DragonFly already bests some of the so called “commercial” systems in many areas!
Justin, what major work has Dillon been doing on the kernel lately?
Probably best to checkout the commit log to get an idea of what Dillon has done https://gitweb.dragonflybsd.org/dragonfly.git
Like what Anonymous said – but even more specific:
https://gitweb.dragonflybsd.org/dragonfly.git/search?s=Matthew+Dillon;st=author
@justin, even though I read through some of the commit log for Dillon, it’s hard to get a sense of what’s the big picture he’s working on.
It’s almost like each commit log is a single tree in a big forest. Yes, I can look at a few trees but that gives me no sense of what the forest is that’s being built.
As such, what are some major overarching work currently being developed?
Hammer2
Anything else?
When are these major features expected to be apart of which release? E.g. Hammer2 = release 5.0? Etc
Ask Matt directly – he’s usually on EFNet #dragonflybsd during west coast afternoon to evening hours.
It would be really great to know which release might have HAMMER2. I am really interested in using it.
I looked for dragonflybsd channel on EFNet but could not find it
I looked for dragonflybsd channel on EFNet but could not find it
I’ve been idling in #dragonflybsd for several months, so it’s definitely there.
@justin,
Re: #dragonflybsd channel. As you’ll see from last night, I posted a few questions about roadmap and major features … and no one responded.
Kind of disheartening
That roadmap is something 1 person wrote … 2 years ago? Probably should be deleted since it’s setting up a false expectation.
Like I said before, it’s an all-volunteer project with no commercial pressure. People work on what they want to do, so a roadmap isn’t going to apply because it’s not centrally planned.
@Justin
Is there any intent by dragonfly developers to push for wider adoption of usage by the community?
If so, having better communication with the community on roadmap and planned features is needed.
Honestly, besides your website – I don’t feel like anyone else talks about dragonfly.
Which is sad because I’d love for more people to use the OS
Maybe asked a different why, is dragonfly intended to be just a hobby OS or is there aspirations for companies to adopt is and use it in production!
+ 1 is dragonfly a hobby OS or meant to be used in production systems?
I’ve long since wondered this as well. Seems like it has so much potential but no clear communication of its intents.
+ 1 JakobT & Ben
Matthew Dillon has said essentially it’s a research OS. This is an arbitrary/pointless distinction, really. Linux was an academic exercise until it was picked up by commercial companies, at which point it was I guess a “production system”. I can tell you it’s production or hobby or research or whatever, but the code’s still the same high quality. The only way to encourage it is to use it.
As for people talking about stuff… Lemme put this in perspective: I’ve been doing this for more than a decade. We now have BSDNow, and some other sources out there talking about BSD, and some active user groups in the US. 5 years ago – there were none. For any BSD. This is pretty good, considering the New Hotness is DevOps and containers and doing your best to make systems irrelevant.
Not to hate on BSDNow because the two guys obviously spend a lot of their own time broadcasting the show and they are dedicated … but
1. The guys are really hard to watch. They are so scatter brained and the content is so little. Many times they just broadcast them essentially reading a blog article they came across
2. Many times the content isn’t even bsd related
I’m not going to hate on anything BSDNow people have been doing – it’s unprecedented: producing weekly BSD content and keeping on doing it. Very few other people (me, BSDTalk, Dan Langille, Michael W. Lucas) have managed to consistently output BSD material in the long run – and BSDNow is producing video, which is a lot harder than my few-sentences-of-text dailies.
I’m not arguing against your opinion; it’s valid because it’s your own; but I want the scale of what they are doing to be noticed.
@JakobT
I’ve learned quite a bit about the project’s goals by idling around on the IRC channel. A lot of immediate and near term stuff is discussed by the developers, but every now and then there are pretty good discussions about longer term goals. I remember some time ago reading that Dillon said that he was pretty happy with the state of the DragonFlyBSD kernel. Much of what seems to be happening now is squeezing out more performance from already mature subsystems (SMP operations for example), fixing bugs and now heavy work is being done on the DragonFlyBSD vkernel. Dports is working beautifully and a lot of previously non-building ports have been fixed. UEFI booting is now a go, clang support seems to be not too far away.
HAMMER2 is documented here: https://gitweb.dragonflybsd.org/dragonfly.git/blob_plain/HEAD:/sys/vfs/hammer2/DESIGN I am pretty sure I am correct in saying that the entire *BSD community is hoping that Dillon can pull this one off. We are rooting for you!
You can also get a sense of the project’s longer term goals by seeing what they wish to include in the project. For that the Google Summer of Code Project suggestion list is informative: https://www.dragonflybsd.org/docs/developer/gsocprojectspage/ note that many of the suggested items are very serious projects in and of themselves. DragonFlyBSD is definitely not a hobby OS. Perhaps in the true spirit of open source computing the project is essentially defaulting to “look, here is the code, do what you want with it”.
I can certainly see the commercial appeal of using DragonFlyBSD as a cloud computing backend once HAMMER2 is production ready. Perhaps even a NAS or SAN product as well. There are a number of groups that have spent a ton of money trying to make ZFS clusterable / high-availability with varying levels of success. HAMMER2 will do this out of the box as well as have machine-level and network-level redundancy.
Time will tell of course, but I’d say that their unspoken road map is pretty clear.
With regard to the “production system” thing, I can tell you DragonFly is pretty much used in production/money making roles even if it is not widely publicized.
Heck, my company has been using it to manage invoices for close to 10 years now…
Is there 11n support for athn in the next release?
Does HAMMER2 have a chance of being ported to other BSDs? I remember this being difficult with HAMMER1 without major parts of the Dragonfly kernel needing to be implemented as well. But then again, FreeBSD basically did that with ZFS and the Solaris kernel.
The ath(4) man page mentions 11n as supported by the AR5416 ath chipsets, and those are a supported chipset – so it looks like yes. Note that I don’t have anything with this chipset to find out.
https://leaf.dragonflybsd.org/cgi/web-man?command=ath§ion=ANY
sephe in #dragonflybsd (Sepherosa Ziehau) is the one who knows.
@Francois Tigeot
Do you have any current benchmarks of Dragonfly running Postgres vs FreeBSD 11?
I think this is the benchmark Ricky is referring too from 2014.
https://www.pgcon.org/2014/schedule/attachments/309_PG_as_bench_mark.pdf
I too would love to see an update of this benchmark.
+1 for someone to run current Postgres benchmark and compare against FreeBSD
If it is not too much trouble, can Linux be added to that Postgres benchmark?
I have a single answer that works for all of this: you all have the same access to the same tools as I or any DragonFly developer has. If you want benchmarks or release plans or something else… get started.
@Ricky, @Anonymous: I’d love to but I’ll need two things for that to happen: access to a similar dual-xeon box and free time. Lots of free time.
@Francios Tigeot
If I donated the dedicated server below for a month, would that help you.
https://www.hetzner.de/gr/hosting/produkte_rootserver/dx291
@Andrew: it could. I’ll need to be able to quickly change operating systems, which I did by physically plugging/unplugging SATA disks last time.
Francois
Why do you need to “quickly change” the OS?
As opposed to simply separate OS installed on physically different harddrivrle. And then shuting down and rebooting into the other environment?
That’s more or less what I meant; it’s difficult to know beforehand if a remote server will be easily manageable wrt system installations/changing boot drives etc…