Sun, get real! solaris is dead, move on.
Yesterday, July 11, was one one of those Sun days… Lots of hardware announcements, good press, and even positive notes from stock analysts, with a long time critic, says CNET, taking time to upgrade SUNW stock rating. Lets be fair, the new gear is cool, and deserves all the positive quotes it is getting but good hardware wasn ‘t never Sun’s weeakest spot – software is.
And yesterday no one, from Sun, talked seriously about software. A good idea since Sun’s software side is seriously messed. Java ? A damn good idea that stopped in time. It’s issue isn’t opensource or not to opensource it, it’s complexity. Java, when it come, was simpler and more flexible than anything else, today is a gigantic soup of acronyms. It ’s still good, and usefull, but someday someone at Sun forget what was java about when it was born. One takes a look, now, at python, a look at ruby, and ruby and rails (take this post i got from… yes, from planet solaris), and feels the same today, that feeled once – those tecnologies are now to java, what java was once to C++. And that is saying almost everything. Will not replace java, as it hasn ‘t replaced C++ but then, that wasn’t the idea. The idea was to make a paradigm shift – write once, run everywhere. Now the idea was upgraded to things like K(eep)I(t)S(imple)S, D(ont)R(epeat)Y(ourself), convention over expression. In short, the revolution is going, with Sun sitting in the sand. Embedded Java? Sure, as cool as it gets, but even python was recently ported to Symbian…
Then, we have gigantic and ever expanding Sun ’s ’software stack’. Sun should take a look about what it is doing in hardware, commodizing what has to be commodized, and differiantiating where trully matters, which is, by the way, exactly the same thing that … IBM is doing in it’s software business. But no, Sun insists in having everything and the kitchen sink, doesn ‘t seeming to have the minimal idea about what is essencial, and what is accessory, about what they are really good, and about were they are simply a burden. Take a look. They fail to market decently trully amazing products (like they ‘re SunRay stuff) yet they insist in market and invest in stuff that doesn ‘t makes any sense, for them. Mail software ? Do they really think they have something comparable, for instance, to Zimbra ? Web Servers ? Does Sun really think that investing against apache (instead of building on top of it, as IBM does) make any strategic sense ? Java IDEs ? Is really that wise wasting time to compete against eclipse ? SunStudio ? Does the world needs another general purpose compiler ?
And then we have solaris, the alleged crown jewell, opensourced and everything. So what ? Sun seems to forget a basic, and elementar, fact – numbers changed from ten years ago. Demographics too. Then, Solaris was the de facto standard in the data center, almost everybody, from school up, had grown with Sun gear, and solaris, close to his mind. Today, things are trully different, like it or not, its either Linux or Microsoft, with the remaining Unixes relegated to niches. Solaris only trully appeals to a few, and Sun, again, is making all the moves, to ensure that it doesn ‘t succeed. While selling boxes that run linux as good or better than anything else, it constantly bashes it, making it a second class citizen in it’s software world. Solaris is supposed to be the best thing, the crown jewell… Really ? Then, how to explain the dumb proposal do HP, made recently, to dumb it together with HP-UX, and make something new ? Is that the kind of assuring news that will get people to left Linux and embrace Solaris ?
More, for all the cool stuff it has, Solaris is in some very critical areas simply outdated. Take software packaging, a gigantic pre-historic mess. Take Xen support, a critical item, where Solaris is at least 12/18 months behind Linux. And take some gigantic misteps. Sure, zones are fine but, was that really bright wast time and resources and add a Linux personallity to them (supposedly to run RH (!) certified SW) and delay xen? ZFS is fine too, and dtrace, but does that justify the switch ? Won ‘t Sun gain much more if it made ZFS to run on top of Linux ? Take the acclaimmed niagara servers, after all they serve more http requests under linux, than under… solaris. Take even Nexenta, basically Ubuntu and debian repackaged on top of the solaris kernel, done well outside Sun, and we start to see that something is deeply wrong in Sun’s priorities. After all, they are no more the ecosystem, they are part of something larger, much larger. Worst , the excessive and melodramatic overhyping of solaris ends damaging the HW side of the business, specially the x86 side. I know lots of people who distrusts Sun’s Opteron boxes (trully world class) simply because they think that Linux won ‘t run fully on them. Not true, but perception matters.
Sun needs to get humble again, needs to get its priorities straigh, if it wants to ever again become trully relevant in the software world, not only opensource…
Trackbacks
Use this link to trackback from your own site.
totally agree. Great post
While, I can agree with some points. Sun has over the years, has got somethings wrong. Open Solaris is definitely not one of them.
One thing I certainly dont want the to do is to become just another Linux bandwagon junky. That would be their death.
If you dont think much of zones, zfs, and dtrace to use Solaris then that is your choice. I can see some useful things on Linux when I use it. I would have to say that I am using less and less. It lacks dtrace, zfs, zones, smf, good resource management, a stable kernel abi.
What would Sun gain in putting ZFS, Zones on Linux rather then Solaris? They would simply follow SGI.
I certainly agree, they have failed to capitalize on the Sun Ray front. They absolutely rock.
Doug, to me, atm, the killer feature is hw virtualization, more exactly xen. To me, and to every other *nix admin i know, in the field. Solaris lacks it, and will lack for another year at least. Sure, dtrace, zfs, zones, smf, good resource management, a stable kernel abi are good to have, but for the kind of workloads i usually do, xen is simply more important…
I think that a lot of this depends on your perspective. In the environment I work in dtrace is a major selling point for me. The ability to debug a system that’s in production without a major performance impact? That’s worth its weight in gold to me. Commands like prtdiag, not available (or useless) in non-Sparc hardware, are also very important.
Solaris is meant to be a bullet-proof, highly scalable OS, and I think it succeeds very well there. I personally prefer Linux (I think it’s easier to use) but if I need something that absolutely needs to stay up and is secure to boot, Solaris is the way to go. Virtualization is an important feature, but I, personally, am not convinced that it’s a must-have app. There are some places where it’s usefull, but many places where it’s not.
Should everything be Linux now? I think there needs to be more variety and competition out there.
OpenSolaris and the *BSDs can cross-pollinate their source code, you know. I see the future of UNIX as probably being the two primary “legacy” (though this is debatle) *nix’s, AIX and HP-UX, along with the open source *nix’s of BSD/Solaris and Linux.
Must open source software can run on Solaris anyway. And it’s not hard to repackage closed source for variety POSIX-compliant systems. Sun can customized Solaris to fit its hardware like a glove, so that can be a major selling point for it.
Why throw out such excellent technology anyway? Does everything have to be a computing monoculture just because?
Also, Linux and Solaris have different goals: Linux is not meant to be as rock-solid as some claim. It’s really suppose to just be “good enough” and be feature-rich and extremely multipurpose.
Solaris is designed for web application services and databases. It goes far beyond just “good enough” however into the “excellent” category.
alucinor… I ‘ll not rebate your point. I even mentioned Nexenta, remember ? I’ll simply reinforce that what Sun is doing now is not what you ‘re proposing.
As a side note ‘Linux’ is much more than a kernel, is a concept, and an ecosystem. Sun would gain **much** more if it could fuse it’s own libc with the standard – glibc, and if diferent kernels (*nix, solaris, whatever) could really share a common userland API (more abrangent than POSIX…).
The story isn’t about what if scenarios, it’s about facts. I finish noting that you recognize that Linux and Solaris have diferent goals. Did ever Sun strategists recognize(d) that ?
Linux is great for what it does. But it isn’t Solaris, which is far from dead. In a big organization, where your system rebooting will cost millions of dollars in the couple minutes it takes to come back up, you don’t put critical infrastructure on Linux.
Solaris has a very good role in large enterprises. I don’t think it should be running small webservers any more than you do, and the places where Xen is likely to be “incredibly useful” are not the same places where Solaris is likely to be “incredibly useful”. We’ve got hundreds of servers (Solaris, AIX, Linux) and we don’t use Xen on ANY of them.
Pat… Ok, you don ‘t have a need for virtualization. Great. I, and the vast majority, don ‘t have as that money, so we ‘ll need it. And even if we had, right now one of the great ways of achieving greater RAS is… guess what – Virtualization. Or do you think that SUN put a hw supervysor in the niagara chips (which sadly no OS is using) just by accident ?
The only thing i feel about sun’s software is slow. It’s IDE, app server, or even Java itself, are way slower than its competitors. In my opinion, most of its works are over engineered.
I’ve found solaris to be sub-par. It’ll boot a sun workstation, but that’s about all I’ve found it good for. Stuff in the os has been so neglected that tcp wrappers is a 3rd party add-on (or was in Solaris 9 anyway). The author of the article isn’t familiar with java. I have a feeling he thinks its something for making pretty web pages because he’s generalized it with ror and others. Java was not intended, originally, to be a Flash replacement. It just evolved that way through carless programming, and that has developed a widely believed stigma. Java was originally intended to be a cross-platform byte-code interpreter – with applet functionality. Many columnists do not know this, and hence are misinformed. Java code you write on one platform will run on any other platform that has the same version of JVM. This means you can develop a java application on x86 Windows, and run the same application on a RISC based ARM processor (if the JVM is there). The overhead the interpreter needs is staggering – ram and cpu usage is very heavy – especially if any Swing is needed. Threads consume a large amount of memory as well, and grow exponentially. Java isn’t right for every application, but when your hardware costs less than your development costs, and you need multi-platform support, Java is what the others are not.
Linux SUCKS! There I said it! No kernel binary compatibilty, no kernel API for modules and stuff developed by school kids is crap!.
Solaris rocks and is the best.
So what’s my point? …. What’s your point?
> The author of the article isn’t familiar with java. I have a feeling
> he thinks its something for making pretty web pages because
> he’s generalized it with ror and others.
I’m no Ruby programmer, but I can testify that Python has been better for my write-once-run-anywhere ambitions than Java ever was.
For starters, I can distribute the python runtime with my program without too much additional disk space or licensing issues*
I don’t know much about Solaris, but after trying to install it in VMWare, and it complaining about needing more than 170MB RAM, I stopped the installation, and have gone back to ignoring it.
> Solaris rocks and is the best.
> So what’s my point? …. What’s your point?
My point is, Sun needs to step back and look at their software offerings, and how its damaging their business.
And you’re right, “Solaris” is a pretty word. My gf says it sounds nicer than Linux.
[...] The big news of the day isn ‘t that Sun is freeing java under GPL, the big news is that they are considering to do the same with… Opensolaris. That is, obviously, the right thing to do. Not only Linux would benefit from some cool stuff Solaris has (in one word – ZFS), but Sun would benefit too, and a lot. After all, and independently (me) of considering that in the long term there are no space for OpenSolaris, as it stands today, as a general purpose end-to-end mass market operating system, there are too many stuff that would only benefit all sides if developed jointly and in parallel, avoinding unnecessary effort duplication. Stuff like Xen, device drivers, even filesystems, hal, acpi, lots of APIs, etc, would gain a lot if developed/released/synced a layer above the OS (as Xorg proved that it is possible, and scalable). For the moment, and under CDDL, the ‘opensourcing’ of Solaris only appealed to those already in the Solaris camp (mainly inside Sun) having little or no impact in the general *nix landscape. If Sun truly believes they have a superior OS, as they say, they should have no fear going the GPL route because it will be the linux features that would land in Solaris and not the other way (emptying solaris’ appeal)… [...]
You are so misguided in you thoughts about Sun. What the hell do you think has been the back-bone of the internet?..Unix and Sun! try placing a bunch of linux crap servers in a ISP email environment a see what happens…they fail! besides linux is just a variant of v4 Unix not to mention pieces of other Unix variants. Besides, the support and documentation is crap… basically linux is a “hobbyist” piece of software.
So what are the things that Solaris needs to provide to win back the hearts and minds of developers in terms of tools, features, protocols etc.
[...] the past i ranted a bit or two about Sun Microsystems. Today i feel the need to do it again. Make no mistake – I like them, [...]