Solaris has gone back and forth between fully commercial, to free + free updates, to commercial, to free without updates, to now commercial again.
In my experience using solaris at the bottom-end has been kind of a non-starter for a decade now because Sun yo-yoed the licensing enough to remind people they shouldn't use it.
Honestly, good riddance. Oracle can go figure out what they're trying to do and this will motivate the open source OS hacker community to get the relevant features of Solaris implemented (or surpassed) in other OSes.
SystemTap (the kernel probing tool build into Linux) is coming along pretty well - I needed to create something to alert me when arbitrary apps set TCP_NODELAY last week, spend a little while looking at SystemTap, made a simple script that probes at tcp.setsockopt(), got the info I needed, and printed it presentably.
True. I believe they're adding providers to the major VMs though (I mainly work in Python so I focus on the cPython VM, but I have heard of some early JVM introspection work too).
Yup, DTrace and ZFS are pretty much the only things Solaris brings to the table above what Linux offers.
Interestingly I've never come across any pure Solaris shops that run ZFS, I suppose because if you're conservative enough only to use Solaris, you're too conservative to trust ZFS.
Solaris zones are awful, and a good example of where Linux is far superior to Solaris. The creators of Zones seemed to have user-mode-Linux as their model, and it shares most of the annoyances of that too. It's not virtualization, and it's not a jail - it's a middle ground which is good for neither.
Agree with you to some extent on stability - I've seen Solaris boxes with a load of 300+ still up and running although unusably slow, where Linux would almost certainly have died from resource starvation. In general use, both Linux and Solaris are pretty damn reliable though.
Whats so bad about zones?
It's clear that they are no virtualization. As you said it's a middleground, more capable as jails but not virtual, which is fine if you don't need it?
What does that get me over Upstart or the other event-driven init systems?
(This is a question of genuine curiosity. SMF has led to many headaches when I've had to deal with it, simply because it didn't seem to comply with normal conventions like returning nonzero for failure or actually printing what goes wrong to the console.)
Call me a BSD bigot (even though I use Linux/OS X at work) but I really fail to see any rational reason to use Solaris at this point.
We have a Solaris database + filestore at work, mostly for Z-raid, but now that FreeBSD has ZFS, when it gets EOL'd (with a vicious grin on my face) it's going to be a BSD box that replaces it.
Sun was embracing Open Source to some degree - though they were always a little schizophrenic on that policy, and never really held to a clear position for about 3-5 years in the middle - but the last few years they seemed to "get it". While this is a minor change (it doesn't appear to affect OpenSolaris) it does seem to be indicative of Oracle taking a few (small) steps away from that position.
But, on the flip side (and I say this as someone who learned Unix on SunOS 2.6 in 1998 - I still type ps -ef, and as a Joyent user, though tcsh has given way to bash in my finger memory) - Solaris really is an edge operating system for the masses, somewhat more akin to AIX. Pretty much the common (server) OS for the masses is Linux, with Open/Free/NetBSD making up the majority of those not going with Windows.
The only people I know using Solaris these days (and this is just an anecdote, so be gentle :-) are people running Oracle or other environments that need to be enterprise robust on vertically scaled system, people using sparc gear, or people who have a legacy investment in Solaris. This is sad - as few as five years ago, Solaris was still a popular choice for new companies - the OS was just so rock solid stable.
I'd love to hear a YC survey on OS choices for server backends - my bet is 50%+ on Linux/BSD, less than 10% Solaris.
I'm wondering what percentage of the remaining 40% is Windows, and if AIX ever gets a vote. I'm presuming HPUX is not chosen for new companies any more, and I fervently pray that everyone steers clear of SCO.
I'm willing to bet that your percentages are off. I would say at least 85% of the folks around here use Linux on their servers, with a 1-5% using a BSD and the 1-14% using Windows. Solaris (read OpenSolaris) filling in the gaps within the fluctuations above.. which is a shame, there is some really cool technology in OpenSolaris/Solaris 10.
I think your predictions are pretty accurate. I would use Solaris because it does have a lot of cool features (Solaris zones being one of the most interesting to me), but I'm unsure about the licensing situation, so I've held off for now.
The thing that has always impressed me about Solaris is its well-roundedness. Not the user-environment, which compares badly to typical linux and BSD setups (svrV vs BSD kludge; standard install lacks lots of useful tools; dev tools and libraries less likely to have support). Rather, the mechanisms that bind it to the hardware, and allow you to do types of control and monitoring that aren't feasible with linux/freebsd stuff. Solaris has had that close-to-the-hardware advantage.
Personally I've migrated from FreeBSD to Ubuntu for my at home server tasks. At work, I've seen a variety of deployments, never one of Solaris.
I'd say 30% Windows, 60% Linux, 10% FreeBSD (instance installs (host and vm's), not including embedded systems like routers) is what I've seen over the past 5 years are work.
Of Linux I've seen SUSE, RHEL, and Ubuntu widely used. I've seen deliveries of Sun Servers with Solaris installed that were wiped to be replaced with Linux.
On the Windows side, I've seen environments where it's Windows Server heavy require drastically more resources to maintain, but that's another conversation.
Of course this is all anecdotal as I've also never seen anyone deploy OSX Server in any capacity :)
What I really think it is, is people use what they are familiar with or what they can get support for. Solaris has disappeared mostly from the desktop, so people are less familiar.
Hell the last time I tried to experiment with OpenSolaris I couldn't get it booted in VMware (graphical corruption), so I never had a chance to get familar with it myself.
To bad too, DTrace and ZFS always sounded fantastic.
I've also never seen anyone deploy OSX Server in any capacity
I've done it, but it wasn't very pretty. Well, I mean the admin tools are a pretty GUI, but underneath it is all a BSD, but instead of being able to upgrade components, you're tied to a vendor that doesn't really update things that way. For example, I had 2 G5 Xserves running 10.3 (Panther) and one Intel Xserve running 10.4 (Tiger). Unfortunately upgrading the version of Python or MySQL on the systems was a royal pain in the ass. Custom configurations is also odd, because they do a lot of custom stuff to the point that it isn't quite a *nix. But if you run off the shelf things designed for a Mac Server or need to use one of the Core Media frameworks they would be good. However, for typical web stuff, they really aren't that good of a server. Great hardware though... I'd love to rip out the Mac OS and put Linux of them.
To put this back into context for the post, OSX Server is probably deployed less often than Solaris.
My last job, about 80% of the systems were Solaris, 15% AIX and 5% Linux/Windows. Granted, this was a large financial institution.
The place prior to that was about 90% OpenBSD, 10% Arch Linux. A startup.
And before that, probably about 25% Linux, 40% Windows, and the rest were a good balance of AIX and Solaris, at a community college. For what it's worth, the college was using Solaris for students to play on, and Sun curriculum for "UNIX" courses, with the thought that students could go get Sun Certified when they're done with the class; and if they did so, wouldn't need to take the final exam. I'm not sure that'll be an option in the future.
I'm using Nexenta (Debian/Ubuntu-based userland on OpenSolaris kernel) for my home NAS, primarily because of ZFS. Works well enough, but of course it's not a production server.
Personally, I use Linux servers, but at work we still buy and use Solaris 10 on Sun boxes. Yes, 90% of the databases we access are Oracle. Yes it's very enterprisey.
No, not directly. But they're trying to squeeze as much money as possible out of hardware and licensing by limiting users' ability to run Solaris for free on beige boxes, and that does seem to tilt the odds in the open question of whether Oracle will continue to let OpenSolaris thrive, or whether they will marginalize it to maximize short-term profit.
I say short-term profit because, in taking this course, Oracle seems to forget that many of the part-time tinkerers and educational or small business users whom they've just priced out of Solaris will be the ones making "enterprise" purchasing decisions in the future. This revised licensing agreement ensures that such users will be more comfortable and familiar with the competition instead--Linux, BSD, even Windows (Microsoft has done a fantastic job ensuring that their software is affordable in the education and small business sectors, while simultaneously maintaining a healthy profit margin in enterprise sales).
My biggest fear is that some Oracle exec seems this a a way to meet his revenue target at the expense of making Solaris irrelevant and risking the same for OpenSolaris.
But OpenSolaris is still open source. If the community believes OpenSolaris is such an important asset to lose I am sure they can pick up the development if they want. And we already have the most important OpenSolaris technologies in FreeBSD so it wouldn't be that big of a deal anyways.
My guess: they are about to ship ZFS deduplication in released version of Solaris (right now dedupe is only in OpenSolaris), and realize that they can bump up their numbers, at least temporarily, by trying to charge for it.
In my experience using solaris at the bottom-end has been kind of a non-starter for a decade now because Sun yo-yoed the licensing enough to remind people they shouldn't use it.
Honestly, good riddance. Oracle can go figure out what they're trying to do and this will motivate the open source OS hacker community to get the relevant features of Solaris implemented (or surpassed) in other OSes.