Great work Colin! I'm definitely going to jump on the free usage tier and experiment with it. I've been holding out, entirely, for FreeBSD to come to EC2.
Should I be able to use FreeBSD on Spot Instances? I'm more interested in this for the future, when larger instance types will be available, but still.
Looks like I can't use any of the Community AMIs on Spot Instances at the moment.
Update: Yes, I can run ami-c01aeca9 on us-east as a Spot Instance! Though any other Community AMI I tried (e.g. Turnkey Linux AMIs) refused to work for me. Any comment on this is still welcome.
I've been looking for a way to move off of Joyent shared (solaris) and onto EC2 with FreeBSD for quite some time. This is awesome news. I honestly didn't know if this day would ever come.
One could turn the tables around and wonder why people would chose Linux over BSD.
To me the biggest advantage is that BSD is fully integrated (it's a full system, not just a kernel) and there are compatibility guarantees. For example a binary built for 8.0 will stay compatible with the whole 8.x series.
> BSD is fully integrated (it's a full system, not just a kernel) and there are compatibility guarantees.
That's true of Debian, too, though. We can disagree on which project does a better job at it, but they're both full systems, compatibility tested and with long-term support for stable releases.
FreeBSD and Debian are integrated in different ways. In Debian, you have a single packaging system which integrates kernel, libraries, and applications, but each of those is developed independently. In FreeBSD, the kernel and base system are developed together -- but we have two packaging systems, one for the base system and one for 3rd party applications.
Basically, Debian integrates everything, but only loosely, while FreeBSD has tighter integration, but only of a smaller subset.
It's not true in the same way. Debian and GNU are two separate projects for instance, but Debian relies on the GNU userland. Same goes with the kernel obviously, and numerous other projects. The BSD projects however (for the most part) do all their own code, all under a single source tree.
The merits of each method are up for debate, however they are pretty distinct.
The extent of integration imo depends on what kind of system you're running. For some systems, especially minimalist ones, I agree the BSDs tend to be more integrated, though they do still heavily depend on third-party stuff (the compiler, whether LLVM or GCC, being a big one).
However, once you start adding some additional packages, Debian tends to do the integration-testing more pervasively across the entire userland, "Debianizing" packages so that they all store things in the standard locations, are started/stopped using the same methods, follow the same configuration policies, etc. Once you get out into ports, the BSDs tend to give you more of an as-is set of 3rd-party packages, and are pretty lax about integration testing.
> (the compiler, whether LLVM or GCC, being a big one).
Although I think the effort is misguided, they are working pretty hard on that one. There is some considerable effort going on to get pcc as the default compiler for the BSDs, specifically so that they don't have to rely on another upstream (and so they don't have to use a GPL'd compiler).
However, once you start considering obviously 3rd party packages, then not even windows could be considered a single project piece of software.
Might also be worth noting that Debian's disregard of the upstream (compared to say Fedora, which makes an effort to get patches into the upstream before using them), has gotten them into trouble on a few occasions.
The most prominent in recent time being their absolutely braindead patching of OpenSSL which dramatically reduced the entropy in key generation, which would have never made it through the peer review process of the upstream... Debian likes to think that they are the upstream of the projects they use, but unlike the BSDs they really are not.
I'd recommend waiting for some more bugs to be worked out before you start using this in production -- but please feel free to start launching instances and helping me find said bugs. :-)
Could you explain in layman's terms, where did the difficulties lie in using FreeBSD on EC2? Since I think I heard an announcement that FreeBSD 8 supports Xen domU, though I can't find any information to this effect now and FreeBSD Handbook doesn't mention it either: http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/vi...
I don't mind NDAs as long as they're reasonable and I get something out of signing them. In general I'd rather know things and not be able to talk about them than not know in the first place.
As FreeBSD security officer I handle 'do not disclose until date X' information on a regular basis.
Yes, unlike the OpenBSD guys, the FreeBSD guys don't mind being under NDA if it means they can release support for hardware sooner rather than later. Look at Sam Leffler and the ath driver (although he is no longer under NDA)
FreeBSD is kind of another beast at this point in time, NetBSD support for xen is actually okay and it is pretty trivial to get it working, freebsd not so much.
Hopefully cperciva's work will pay off soon for a stable freebsd on Xen as I know there are plenty of users, me included, that would love this option.
Edit: and that linode post on how to get it running looks a bit dated
Thank you. I've heard fantastic things about FreeBSD for a while now, and now in combination with the micro instances, I can learn and experiment. My EC2 playground is already ported over to FreeBSD, portsnap updated, extracted with a fresh nginx install.
It's been a pleasure to use thus far. Thank you, again, Colin.
Oh, this just made my week. I can't wait to get my free tiny instance and start playing with my shiny new toy. I cut my Unix teeth on FreeBSD, and hearing that it's well on it's way into the cloud...well, frankly, I just squeeed a bit inside...
Why do you need this in EU-West? I'm planning on building AMIs for the other regions once this is more production-ready, but at the moment I can't see why it would matter which region it's in.
Dying? That may be true of the other BSD's but FreeBSD appears to be thriving; some very big companies, universities, &c... use and rely on FreeBSD. OpenBSD appears to be alive and well too, but from my perspective FreeBSD tends to have the spotlight.
Definitely not dying. In fact I think FreeBSD will just increase in popularity in the years to come. It had a bit of a dip in the last 3-4 years, but even from that it has shown how well the project holds up.
