GhostBSD and OpenRC

[NevilleGoddard]

Hi! I hope I'm not offending anybody by suggesting this.
I watched a video on YouTube called "Kris Moore - Unveiling TrueOS,MeetBSD2016" in which he mentions he has successfully ported OpenRC to TrueOS and this has significantly improved boot up times. I myself tried TrueOS on my old laptop and I can confirm that boot up times went from about 120 seconds (with the old init system) to 70 seconds (with OpenRC). GhostBSD with UFS would boot up much faster than this. Moore says that OpenRC was developed originally by a NetBSD developer and it was easy to make it work in TrueOS. Would it be possible for GhostBSD to use OpenRC? It would almost halve boot up times.

Also, will GhostBSD11 also have an Xfce version?

Thanks.

[ASX]

Hi! I hope I'm not offending anybody by suggesting this.
I watched a video on YouTube called "Kris Moore - Unveiling TrueOS,MeetBSD2016" in which he mentions he has successfully ported OpenRC to TrueOS and this has significantly improved boot up times. I myself tried TrueOS on my old laptop and I can confirm that boot up times went from about 120 seconds (with the old init system) to 70 seconds (with OpenRC).

How did you compared the two options ? You deinstalled OpenRC and reinstalled the previous init ?

GhostBSD with UFS would boot up much faster than this.

I have to read this as UFS will be faster than ZFS .... which is not, in my experience.

Moore says that OpenRC was developed originally by a NetBSD developer and it was easy to make it work in TrueOS. Would it be possible for GhostBSD to use OpenRC? It would almost halve boot up times.

I suppose that yes, it would be possible, but I'm not sure we want go that route. Parallell services startup introduce some problem, especially for LiveDVD where we need a specific sequence to start successfully.

Also, will GhostBSD11 also have an Xfce version?

Thanks.

We have the XFCE version and is not going away any time soon.

[NevilleGoddard]

Hi, ASX,
I compared the two options by remembering the boot time for the old init version of TrueOS which was then PC-BSD. It took almost 2 minutes to boot. I downloaded the latest version of TrueOS which had OpenRC and boot times were much faster. In my experience, boot times with ZFS were slower compared to UFS, but other operations under ZFS were probably just as fast or faster. If you're not going to use OpenRC, no problem as GhostBSD is the best OS out there by far.
Thank you for keeping an Xfce version going for GhostBSD11.
Can you upgrade to GhostBSD11 from Enoch or do you need to delete the old version and do a clean install?

[ASX]

Hi, ASX,
I compared the two options by remembering the boot time for the old init version of TrueOS which was then PC-BSD.

That's what I suspected ... and I'm not sure the two version are exactly comparable.

It took almost 2 minutes to boot. I downloaded the latest version of TrueOS which had OpenRC and boot times were much faster.

I'm glad to read that TrueOS improved the boot time, but again, remain to be seen if it depend from OpenRC.

In my experience, boot times with ZFS were slower compared to UFS, but other operations under ZFS were probably just as fast or faster.

I have done quite a lot of testing (testing the installer) using both UFS and/or ZFS (assuming ZFS on single disk here), and I can tell you there is not a significant difference, although I have found ZFS slow when using 512 byte sector compared to ZFS on 4Kb sectors.

If you're not going to use OpenRC, no problem as GhostBSD is the best OS out there by far.

Certainly we will not if that will prevent us to deploy our LiveDVD, and thanks!

Thank you for keeping an Xfce version going for GhostBSD11.
Can you upgrade to GhostBSD11 from Enoch or do you need to delete the old version and do a clean install?

Yes, GhostBSD try to be fully compatible with FreeBSD, and therefore a similar upgrade procedure is possible, because we are going to setup our repository, we will provide some additional info when we will be ready, namely info to make use of our repo instead of FreeBSD repo.

[ericbsd]

2 minute boot time is really slow, I never achieve that time on Pentium 3, 30 to 40 second is what i had on a Pentium 3 Machine. And about 15 to 20 second on my day to day machine, but since I only care on the performance running GhostBSD, boot time is not issue to me I only boot when I have some update that need to have reboot it is not 20 or 5 second that will make deference to me. Today people want everything now and can't appreciate to wait a bit for something. Not that I am against to having GhostBSD booting a bit faster, but is not what make a Great OS, what make a Great OS is stability, security, performance on run time and friendly UI.

GhostBSD could boot fast, but first it need to boot on live DVD/USB and being able to be installed to start to work on boot time.

[ASX]

Those times reportedly were referred to PCBSD and TrueOS, not GhostBSD.

[kraileth]

Yes, TrueOS has improved boot times considerably with one of the latest updates. I've followed the announcement from Kris Moore that they'd move towards OpenRC with interest and I've tested a new version with OpenRC (after using TrueOS with the old bsd-init for some time). IMO it is a good thing that they've done this and maybe they are blazing a trail for others (us?) to follow. I don't see vanilla FreeBSD heading down that road as such a switch would hardly comply to the POLA (plus FreeBSD is more geared towards servers, anyway).

Truth be told, the folks over at IX were not the first ones in FreeBSD land to experiment with and ship OpenRC. ArchBSD (later renamed to PacBSD) let you choose if you want to use bsd-init or OpenRC when you installed the system. This must have been beginning to mid 2013 if my memory serves me correctly. At a later time they disabled the OpenRC option - there had been problems with it but I don't remember the details. I have no idea though, if they brought it back. I don't follow that OS's development closely anymore and currently the package overview in their website is broken so I can't take a quick look.

Some kind of parallelization of init tasks would certainly be a valuable addition for a desktop oriented project like GhostBSD. However there are quite some takes on creating modern init systems out there (that's probably the one good thing that we have to thank Systemd for - it managed to scare a fair number of people deeply enough to make them work on their own ideas!). I'm not entirely sure if OpenRC is the best match for a FreeBSD based OS. Honestly Kris's announcement surprised me quite a bit. I know that they were playing with relaunchd (now known as jobd) and was expecting to see that introduced at some point in time. It would be very interesting to know what made them settle on OpenRC and favor it over jobd, nosh, runit, ...

But as much as I'd love to have something more modern in GhostBSD I don't think that this will happen anytime soon. There are simply too many other tasks that need to be taken care of. If anybody would volunteer to play around with some of the available solutions however, I'd be very interested to see and discuss the results!

[ericbsd]

kraileth I agree 100% on that! If any one want to see that happen, volunteer surely needed for this.

[ASX]

Some kind of parallelization of init tasks would certainly be a valuable addition for a desktop oriented project like GhostBSD.!

I disagree here. Yes a shorter boot time would be valuable, but the cost would be a loss of compatibility with the FreeBSD ecosystem, something we live of top of, that would be a "high cost" for us.

Additionally, I tried PCBSD in the past, their boot time was long, mostly because they started a lot of services, certainly much more services than FreeBSD or GhostBSD and I doubt parallel startup will be much helpful for systems that start few services.

So, while a new init system cannot be excluded in the future, it will not be any time soon, no matter how many people eventually will join the endevour.

[kraileth]

Some kind of parallelization of init tasks would certainly be a valuable addition for a desktop oriented project like GhostBSD.!

I disagree here. Yes a shorter boot time would be valuable, but the cost would be a loss of compatibility with the FreeBSD ecosystem, something we live of top of, that would be a "high cost" for us.

That's true of course - whenever we decide to depart from the way FreeBSD does things it's never a "do once and be done" effort but in fact means maintaining it on our own. That definitely has to be considered.

Additionally, I tried PCBSD in the past, their boot time was long, mostly because they started a lot of services, certainly much more services than FreeBSD or GhostBSD and I doubt parallel startup will be much helpful for systems that start few services.

Also true. We would not benefit from the move to the same extent as our boot time was quite a bit shorter. Still there are other benefits to using modern replacements. Anyways we all agree that this won't happen anytime soon nor without serious consideration of the options that we have and the implications of those.