OpenSolaris From a Linux Admin and User Perspective 370
MSa writes "How does OpenSolaris, Sun's effort to free its big-iron OS, fare from a Linux user's point of view? Is it merely a passable curiosity right now, or is it truly worth installing? Linux Format takes OpenSolaris for a test drive, examining the similarities and differences between the OS and a typical Linux distro. If you want to sample the mighty ZFS filesystem, OpenSolaris is definitely the way to go."
maybe I should go and play around with this! (Score:5, Interesting)
Ever since the demise of SGI I haven't looked at anything but Linux / BSD, but this makes me wonder if there is maybe life for Solaris after all.
Would be nice if this was more geared towards the server end of things, which is where I would expect you'd deploy solaris much sooner than on the desktop.
Love that they open sourced it... but... (Score:5, Interesting)
I love that Sun open sourced it, however I think that the greatest benifit is not that it's open but that the technologies it offers are available to be reproduced on other nix os's. The biggest issue I have with OpenSolaris is that it's still a single vendor OS. If it forks a few times and actually develops a culture and some competition between vendors than I think it will be more appealing.
That's actually what I hate and love about linux. It's a fragmented and ineffecient community, but because it's fragmented I don't have to worry that it's going away any time soon.
Re:Nexenta (Score:5, Interesting)
Someone asked me this question recently. And for the sixth time I answered with a laundry list of things I didn't like about it. Agian, I was modded Troll for stating I don't like Ubuntu/Kubuntu, and then people got all in a huff.
Like I always say, it is marketed at a certain target audience, and it isn't me.
I suggest that you try out a really good KDE desktop (Arch's KDEMod, Sabayon, openSUSE 11, etc) and the differences should be immediately apparent to you.
As far as whether or not the KDE packages are available in Nexenta, I'm not sure actually.
On VirtualBox, it blows (Score:2, Interesting)
I put up Sun's free VirtualBox VM environment on a MacBook Pro, and both OpenSolaris and Solaris 10 Intel were worthless. Both achieved speeds reminiscent of PearPC.
XP worked OK. Ubuntu was fine.
You'd think if you were going to release a VM, at least you'd make sure your flagship OS would run on it at speeds that would compare favorably to a 20-year-old Amiga.
ZFS rocks (Score:3, Interesting)
Re:From an experienced Admin's perspective (Score:4, Interesting)
Ahem...
Actually the reverse is true...
Since release 10, Solaris has been pretty well stomping the competition in price, performance and throughput. With Solaris supporting pretty much every type of virtualization (including some not offered anywhere else), it's hard to beat.
Solaris as well as OpenSolaris are free, you can download and use either flavor with no cash outlay. Want support? It's cheaper to buy Solaris support from Sun than to buy Linux support from RedHat.
There's no *tying* with Solaris, it's all about choice. I personally choose Solaris over Linux for pretty much any task.
I have a problem with this kind of "open source" (Score:2, Interesting)
Sure, it's Open Source and everything. But the problem is that complex programs like this are often designed with a top-down instead of a bottom-up approach. I mean, this isn't a bazaar, it's a cathedral. Oh, and OpenSolaris is not GPL. *buzz*
There's still one company responsible and only that company will make the changes, because the codebase is so huge that it's a pain in the *** to maintain. Well, eventually many open source projects end up like that, with a huge codebase and with a company. BUT, this wasn't built by the community and it's not likely that it will get enough userbase so that a dev will become interested.
Compare this with the Linux Kernel. Linus message, which was more or less like "Hey guys, I'm making a unix-like kernel, anyone want to join?", was followed by a stampede of developers and testers like you can't imagine.
So, if we want competition, it's very improbable that a cathedral project such as OpenSolaris can compete with Linux, a 100% GPL'ed project built by the whole community.
Maybe it can compete with the bsd clones, but Linux? I don't think so.
Re:From an experienced Admin's perspective (Score:4, Interesting)
Stay away?
What if you need real uptime with a load of 80 on a 32 cpu system? Can Linux handle the load and have years of uptime?
Solaris just works and its made for servers. Linux seems always beta quality with its cutting edgness and is desktop oriented. I would not trust my job to it unless its Debian or RHES which costs $$$ as cutting edge features are not needed on a mission critical server. Solaris scales far better than any BSD or Linux distro out there.
Ask any *real* Unix admin who uses both and more than likely they will say Linux is great for small jobs but Solaris is king for anything else.
Oh and the article discussed a scarcity of third party apps. I found the opposite as most server ERP and database apps are on Solaris than Linux.
Re:ZFS rocks (Score:3, Interesting)
Someone likely could easily issue ZFS as patch directly for the kernel as opposed to in FUSE. The problem is that it would be illegal to use it a such because of the license. Sun has talked about making it all GPLv3 if Linux takes their kernel GPLv3 as well. Linux would gain native ZFS in their kernel, but Sun would gain every device driver from Linux.
The problem is that too many individuals that you can't even contact own individual copyrights in the Linux kernel. It isn't just going to up and change to GPLv3.
Re:From an experienced Admin's perspective (Score:3, Interesting)
However I think this is probably a response to something I've noticed of late, in Asia and South America we don't sell support for Solaris installs any more, they've all moved to Linux, cheaper hardware, a pool of interested young (and therefore cheap) admins, and of course our wonderful software is available on Linux ;)
While Europe and to a lesser extent the US are almost exclusively Solaris (the odd godforsaken HP-UX or AIX box as well to keep me interested) the emerging markets, where the growth is, are moving en-mass to Linux/Open source.
Re:From an experienced Admin's perspective (Score:3, Interesting)
What about hardware support? I keep hearing that the openSolaris kernel just frankly doesn't have many drivers. If I can't install it on my hardware, it isn't doing me any good.
Also, I'd really like to see some basic benchmarks between the kernels. People benchmark the BSD kernel against the Linux kernel on IO, networking, etc.
Show me some quantifiable numbers on openSolaris.
Re:maybe I should go and play around with this! (Score:2, Interesting)
It's a bit late (Score:4, Interesting)
I was a Solaris admin back in the early 90s. I preferred SYSV to BSD for a lot of things. But at this point, I'm just not seeing a compelling reason to go back. Sure, ZFS sounds nice, but I don't really want a system that's slower and more RAM-hungry than Linux, and I don't want an OS with a GPL-incompatible license.
Re:From an experienced Admin's perspective (Score:3, Interesting)
Re:From an experienced Admin's perspective (Score:2, Interesting)
ZFS - Are you really using your server for data storage? SAN or NAS should be a better option depending on your price point
SMF - while nice, i have experienced many different kinds of errors. If one of the dependencies has a problem, the chain breaks and it is a pain to discover the problem.
FMA - if hardware is broken, i would rather the machine be fully broken and out of service. Running on degraded hardware is too much of a risk. if a few bits get switched or some data is not written correctly, you could corrupt data.
Zones - I still have yet to see a reason to use this except for dedicated virtualization servers.
binary compatibility - if you are running custom code without the source, sounds like you have a setup for failure.
rcapd - ulimit can do this per process, and there are also multiple 3rd party open source resource limiters.
processor sets, cpu shares, fair share scheduler - Yes, there are.
I've been in many different places with Solaris, Linux, and a few other random UNIX environments. In almost every case, the Solaris and other random unix environments could be replaced with Linux at 1/10th the cost.
I manage some HPUX servers right now and just the hardware maintenance on each of the servers is more than a few brand new Linux servers each year.
One significant difference... (Score:3, Interesting)
... between Solaris and Linux in the Enterprise is how they react to abuse - namely stupid people running ton of stupid memory hog applications.
Where I work we have Solaris 9 and 10 boxes running literally unattended for 600+ days - they are shared boxes, meaning lot many different applications run on the same OS/FS/Memory/CPUs .
When a particular app goes haywire and starts (many of them are 64-bit apps) - that particular app just gets a NULL back when there is no longer any memory available. The app can hopefully then calm itself down or release some of its caches etc. but the main point is that the other apps are unaffected and so is the OS.
I would not even begin to think how Linux could handle this. It has this insane notion of handing out virtually any amount of memory to applications whether or not there is actually that much memory and swap available. So when things get out of control the ugly and stupid OOM killer thinks it knows better which app to kill - depending on your luck you could end up with sshd or some other good behaving app being killed to give memory to this bad app.
That is scary. Arguably this is all fixable within the applications but ground reality is that App developers are incompetent - at least where I work, they are.
Plus the newer Solaris releases are close to Linux when it comes to performance. So the only incentive to run Linux is hardware support - if you are on non SPARC hardware that is.
Linux hopefully some day will have a good memory management subsystem soon - less fragmentation, more predictability, good accounting etc. But till that time Solaris for the stupid "Enterprise" .
Re:From an experienced Admin's perspective (Score:3, Interesting)
I'm not the original poster, but I agree with him wholeheartedly. In my experience, tasks that can be easily parallelized work well in Solaris (web servers, polling servers, etc). However, tasks that are serial in nature (dealing with a stream of events like IDS or syslog) work *horribly* on Solaris.
When we moved some of our log parsing from Solaris/SPARC hardware to similarly priced Debian/x86 hardware, we expected a 3x improvement in performance just due to the CPU...we actually saw a 10x improvement in performance. We attributed this largely to Solaris' aggressive reservation of CPU cycles for other threads...even when we only had one.
Re:From an experienced Admin's perspective (Score:5, Interesting)
Ok, I'll call you on this one. I'm a SCSA (Solaris Certified System Administrator) and a former Sun SSE. I've worked with Solaris systems going back to 1996 on original Sparcstations (not even Ultrasparcs). I've also worked on Enterprise 10000, 15000, and 25000. We also have a smattering of Sun Fire X4600s, the new AMD Opteron boxes.
I tested Solaris 10 and Red Hat Enterprise Linux 4 (64-bit) on the exact same hardware (X4600), and you know what? Red Hat Enterprise Linux 4 performed better on massive storage I/O than Solaris 10. I have the Oracle ORION benchmarks to prove it. We have over 50 LUNs carved from an HP EVA 8100 and presented to these X4600s, on 4x 4gb HBAs per server. They run Oracle RAC, have 4x quad core AMD Opterons, and Red Hat Enterprise Linux.
Sorry, but Solaris used to be a good OS back in the 90s. They have fallen so far behind it's not even funny. The reality now is that I can run Red Hat and Oracle on a 32 core AMD Opteron box with a hundred LUNs on a fibre channel SAN and it outperforms Solaris now. ZFS is nice, but we use ASM (automated storage management) for Oracle anyway, so ZFS is unnecessary.
Solaris has unfortunately fallen far behind the performance curve, and I doubt they can ever catch up. Your BS about HP not supporting more than 8 LUNs per target is absolutely BS. I can do hundreds of LUNs, and I have systems like that in production.
On support, they all suck. Red Hat, HP, Sun, every one of them sucks. They have all been chasing the bottom and if it ever gets to the point where I'm stumped, they're going to be stumped as well.
Re:Nexenta (Score:4, Interesting)
Re:Nexenta (Score:3, Interesting)
No, I said I've had the argument several times before and don't see the need to repeat it. Just look for previous posts (like the ones I've listed above).
Trolls get off on pissing people off. I'm trying to avoid an argument. How the hell does that make me a troll?
If anything you're trying to bait me, and frankly I could care less.
Re:maybe I should go and play around with this! (Score:4, Interesting)
With regards to ZFS, I dream of there being an equivalent on Linux.
Out of curiosity, what particular features in ZFS do you want on Linux? I mean, it's a large step upwards from solaris disksuite, but compared to the linux device-mapper/filesystem paradigm it's not a particularly large improvement (if it is one at all).
Having actually run ZFS in production, there are some serious drawbacks with the remaining features (copy-on-write fragmentation, problems in SAN environments, etc), that may leave one wishing they'd implemented the ZFS features in a more stackable way so you could easily discard inappropriate layers and features.
Mixed Feelings (Score:3, Interesting)
OpenSolaris takes quite a bit of time getting used to IMHO coming from FreeBSD, OpenBSD, and many many versions of Linux. I used it because I wanted ZFS, virtulization, and also to try something new.
I did move back to FreeBSD after about a week or so since I thought OpenSolaris brought unnecessary learning curves for someone new. Things like 'ps' being different than every other distro, network interface setup and modification is annoying, the number of programs that you can compile outside of their package manager are slim, and overall not very friendly (I don't want to use the GUI, ever). However, I have 4gb of ram and ZFS really should only be run under 64-bit FreeBSD. Qemu doesn't seem to run, Xen isn't even an option for virtulization and WINE doesn't work under 64-bit (these are the main reasons I bought 4gb of ram in the first place).
ZFS has been running flawlessly on FreeBSD for me thus far, and even the maintainer says he's been using it since her ported it over without a hickup. FreeBSD runs version 6 of ZFS, while OpenSolaris currently runs version 11. It IS true, once you go ZFS you don't go back.
I refuse to run Linux, for personal and limiting reasons, and FreeBSD won't let me virtualize. It seems that in the next few days I'll be biting the bullet and moving back to OpenSolaris. It is very nice that ZFS is seamlessly integrated and snapshots are automatically created when updating the system. This ensures you can easily roll back or boot back into an older install to test different things.
All in all OpenSolaris HAS some potential, but their licensing is very wack and limiting. If Sun wants their OS to evolve and take on more users in the community, the licence will really need to be changed.
Re:md broken? (Score:5, Interesting)
No offense taken, but your first line is a typical response from most linux users -- anything that goes wrong is either an admin issue or a hardware issue.
First, md forgetting drives is not a hardware issue. Linux sees the drive, the serial number of the drive is the same, the hardware does not change, the hardware works. Sometimes, you will boot, and it just loses the configuration, so you reconfigure the array, and wait for it to check everything out. For two hours. At 3am.
Wiping out configuration during upgrades happened for two consecutive releases of the master distribution. Everything is backed up, but 3/4's of the machines didn't boot properly after md was upgraded. Turned out this was a pretty known issue. No one ever thought that people would want things migrated. Everyone seems to have a few hours to manually move arrays over.
Look, I'm all for great, open, free technology. The problem is, most people don't think about the big picture. LVM and MD are fine for personal machines that don't do much more than serve up files, or play music, or what have you. Technology like ZFS is designed to be bulletproof, documented, and it has to be supported. Not only that, but given the right amount of RAM, ZFS can outperform many off the shelf RAID systems, and give you flexibility in mirroring, snapshots, and drive support that LVM cannot possibly compare to.
The only reason ZFS hasn't had much news in Linux land is that it 'wasn't invented here' and it isn't GPL. Last I heard, there was a movement underway to reimplement ZFS under the GPL. I would imagine we'll see something in five years or so.
Clarification (Score:3, Interesting)
Sun has not yet released ZFS an openSolaris under GPLv3, which is the first step.
Next, the Linux kernel would need to be GPLv3.
Linus can relicense all of his contributions to GPLv3, but then the kernel can not include any code currently licensed GPLv2. So actually, every developer who contributed code and maintained copyright on that code would have to be contacted, and all of them would have to agree to relicense the code.
Unlike many other projects where people contributed under "GPLv2 or later", the Linux kernel is basically all "GPLv2 specifically".
It would be a logistic nightmare to relicense the Linux kernel, and many developers have stated they would be opposed to it on principle as well.
Sun could just license ZFS and the openSolaris kernel under GPLv2, but in their eyes it would be effectively giving it away for nothing. The reason they'd consider GPLv3, is to entice the Linux kernel to go the same route, and then both can take from each other.
When Sun discovered that Linux wasn't likely to go GPLv3, they decided not to either.