Blogs

RedMonk

Skip to content

Has Solaris Turned the Corner?

Sooner or later I’ll get around to writing up a more comprehensive State of Solaris post I’ve been meaning to put together, but until then I wanted to share a couple of pieces of feedback I’ve been getting on the recently open sourced operating system. Before that, however, let me recap a couple of personal datapoints to give you some background on where I’m coming from, and disclaim our relationship with Sun up front.

So, to the latter point: Sun’s a customer of ours. Feel free to read into that whatever you like. When it comes to Solaris, my experience has been mixed; while Solaris was a fixture at many of the accounts that I worked in my SI days, I never had any particularly fondness for it. Like the DOS/VSE mainframe I worked off on my first gig, I respected Solaris, but never particularly enjoyed it.

Linux, however, really grabbed me the moment I started using it; it was fresh, it was vibrant, it offered me things I couldn’t get anywhere else, and – most importantly – it had a great community. Nor was I the only one; Red Hat and SuSE alike can tell you tons of stories about the gains they were making at Solaris’ expense (here’s a good one from E*Trade), and the opportunity was compelling enough that even vendors that knew something about operating systems, like IBM, threw their weight behind the little operating system that could.

All in all, the direction of things to come – from where I sat – seemed relatively clear. Windows wasn’t going away, and neither was Linux, but the future for Solaris while not bleak – it’s got a massive presence in some mission critical settings, which guaranteed it’d be around for a while – seemed increasingly niche and high end.

So a couple of years ago, I’m sitting in a briefing at Sun’s Burlington campus, where a couple of the folks explain that they’re doubling down on the next version of their flagship operating system. I expressed some polite skepticism, while giving them credit for the things that enterprises have traditionally appreciated about Solaris – the security, reliability, etc. But more and more I came to doubt this strategy, figuring that competing with an operating system whose costs were ammortized across multiple large institutions (HP, IBM, Novell, Oracle, etc), but more importantly the massive Linux community, was less than ideal (as was the anti-Linux rhetoric, which did considerable damage to Sun’s reputation in the open source world; damage that is only now beginning to be undone). While I believed in the disruptive promise of the Java Enterprise System, I was quite skeptical of the scope of investments in Solaris 10.

Then Sun started trotting out customers. Lots of them. And even being cognizant that these were vendors provided by Sun, the feedback on the operating system was remarkably positive. DTrace, Zones, fault tolerance, SMF: each customer appreciated a different feature, and explained why they couldn’t get said feature in a different operating system – some of which was true, and some of which was not. That got my analyst antennas twitching: it’s rare to hear customers so unanimously positive on a product that you yourself are skeptical of.

And then came the open sourcing. I laugh when I’m asked about the actual opening of the code itself, simply because I can recall talking to a reporter on Day Zero who complained that the community was so much smaller than that behind Linux, to which I replied, “It’s the first day. Maybe it’ll grow, and maybe it won’t, but I don’t think it’s fair to count them out when the doors have been open for a few hours.” Fast forward 41 weeks, and I think it’s safe to say that the OpenSolaris community is thriving. It may not rival the size of the Linux community, but it’s an entity in its own right.

So where am I now? Well, I’m fairly positive on the prospects for Solaris. There are still some real problems that need to be fixed in usability (start with the shell), application availability (Zimbra chose to build on half a dozen operating systems – including OS X – before Solaris, with no S10 build on the immediate horizon – despite the fact that potential customers exist) and package management (the lack of that is why I’ll be running Nexenta), but overall it’s really made remarkable strides in a short period of time. It’s relevant to me as an analyst in a way that it was not 24, or even 12 months ago.

But I’m always keen to see what the delta is between awareness and reality with respect to the adoption and other trends of marketshare products. So I’ve been asking questions about Solaris to most of the folks I meet with for the better part of a year now, and as promised, here are a few of the interesting datapoints:

  • Changing Perceptions:
    I spoke with a vendor in the open source space back in August of last year, and discussed Solaris with them at some length. After relating some of the previously discussed customer anecdotes, I told them that it was worth keeping an eye on. Earlier today, I spoke with that same customer, who confirmed that while they still lack a real Solaris story, their customers are asking for it increasingly of late and they’re considering it. Interestingly, it’s not just Solaris 10, but 8 and 9 as well. It’s unclear if that’s a drag effect of the talk that Solaris 10 has generated, but it was an interesting validation of my theory that folks were paying attention to Solaris.

  • Simpler to Work w/ Than Linux:
    Before Linux advocates get cranky, let me be clear: I’m not talking about technology here. Having used both operating systems, I still believe that despite credible advances in Solaris such as SMF, Linux is easier to use – if for no other reason than better available documentation. Instead, I’m relating what one ISV discussed with me lately, which was that supporting the multiple Linux distros was less than trivial for them. While they officially supported Red Hat and SuSE, and unofficially Debian, it’s a lot of work. Solaris, they told me, represents a potentially winning combination of unfragmented open source. Whether or not that’s a good or bad thing from a user perspective can certainly be debated, but it was interesting feedback nonetheless.

  • Solaris & Web 2.0:
    A rewritten networking stack within an operating system is not the sexiest topic in the world, so it’s not surprising that Solaris work in this area has gone mostly unnoticed, but three separate web oriented developers have talked to me recently about the vastly improved way Solaris handles sockets and connections – critical stuff for folks delivering software over the web. What I’m really curious to see, however, is whether or not ZFS can offer partitioning and distribution advantages to next generation data layers, because most of the Web 2.0 players I speak with have problems here. Joyent’s comments above are an indication that this might be an opportunity.

  • What About the Developers?:
    I wouldn’t be doing my job, of course, if I didn’t pay close attention to what developers are saying about some of the innovations within Solaris. The verdict? A lot of folks are impressed. The Apache folks seem to appreciate Zones, both the TextDrive folks and David Orman could probably be counted as ZFS fans, and DTrace continues to impress the hell out of people that see it (including one of the larger internet firms, from what we hear).

I’m not sure what conclusions you might draw from the above datapoints – which, it should be noted, are just datapoints – but I’m beginning to wonder if Solaris hasn’t turned the corner. Do I believe, as Andreessen apparently does, that Solaris is a better Linux than Linux? Not at all, and indeed, it’s not clear to me how long Solaris will enjoy the technical advantages that it does today.

But considering the above, when Jim Grisanzio asks whether a community is creating a market in this case, I have to concede that at least amongst the folks I’m speaking with, that seems to be the case.

Disclaimer: As mentioned, Sun is a RedMonk customer, as is IBM. E*Trade, Joyent, Red Hat, SuSE, TextDrive, and Zimbra are not, but should be ;)

Categories: Emerging Technologies.

  • http://jroller.com/page/jaimec Jaime Cardoso

    Cool review, I liked it (and, It’s clear that you’re still a bit sceptical, that’s OK, some of us hope to convince you otherwise) but, as (I think) you know, I *really* do NOT agree with you in the package Management arena. Solaris still serves a lot of critical systems where tools like Yast or apt-get are as needed as an hammer in the head.

  • http://www.redmonk.com/jgovernor james Governor

    man… you aren’t tracking developers again are you?

  • Gyro Gearloose

    Interesting Article, but what do you mean by:
    There are still some real problems that need to be fixed in usability [cut some stuff out] and package management (the lack of that is why I’ll be running Nexenta) …

    (Software) Package Management is a part of Solaris since 1995 /1996 (AFAIrecall). (Try commands such as ‘pkkadd’, ‘pkginfo’).
    The only problem is, that many software vendors ignore the mechanism or use it incorrectly. Like other mechanisms & APIs of a rich system like Solaris it takes effort and time, to use them right.
    Have only limited experience with Linux, but the multitude of Package Management solutions, that have emerged will probably not help the situation.

  • Chris Rijk

    I’d say it definitely has turned *a* corner. It’s not that long ago that comments like “Sun/Solaris/SPARC is irrelevant” were relatively noticible. I think that’s pretty much gone away. And I think Solaris is taken a lot more seriously now – particularly on x86, which helps in some subtle ways.

    Customers are starting to use Solaris 10 features. I’ve noticed a couple of examples mentioning Solaris Containers this year. I also remember an anecdote about a particular market segment dominated by two ISVs where one got major DTrace religion and at customer sites is now beating up on the other ISV, which has ignored DTrace.

    Still another corner to go I think.

    In general about installed bases / communities, I would say that there is always some natural attrition – you’ll lose some existing customers every year for whatever reason (the most basic being a customer going bust). So if you don’t get new customers you’re guaranteed to shrink. Until Solaris 10, I think the Solaris community was slowly shrinking – both with attrition and customers jumping. I think the jumping has more or less stopped and there’s enough incoming to cover (probably more than cover) for attrition. However, in terms of actual customers using Solaris for production, though they may well be growth, it’s modest enough that it doesn’t seem to be causing any pain for competitors.

    PS With regards to the Solaris installer, Sun engineers have started a discussion on what they’re working on next and why:
    http://www.opensolaris.org/jive/thread.jspa?threadID=7070&tstart=0
    http://www.opensolaris.org/os/community/install/files/install_strategy.pdf

    Quote from the PDF: In order to gauge our position relative to the competition in the SMB/Developer market, Solaris marketing has performed a comparative analysis of Solaris interactive installation versus the competition. Generally, the results can be summarized thusly: Solaris
    installation is ugly, slow, and difficult.

    Another quote: In terms of ease-of-use, the Solaris installation experience leaves much to be desired. As noted, our own studies show that the experience is not as easy as with our competition. It’s interesting to observe that the core installation experience dates essentially to Solaris 2.1, and has not been especially updated since that time.

    Not that they’re only looking at the competition: One important consideration is that the need here is not just to supply patches – it’s also to supply completely new sets of packages for a particular subsystem, such as Gnome. This requirement is reflected strongly in the Solaris/OpenSolaris development community. Tools such as BFU, Blastwave’s pkg-get, and scripts provided by the various Solaris consolidations such as X server or JDS to update their components, are all evidence of unmet requirements in the core packaging and software maintenance system.

  • http://www.gnusolaris.org erast

    For those who curious what Nexenta means, please read it here: http://www.nexenta.com and http://www.gnusolaris.org.

    Overall great comments!

  • http://phillipfayers.blogspot.com/ Phillip Fayers

    it’s not clear to me how long Solaris will enjoy the technical advantages that it does today

    I have a feeling that Solaris will enjoy technical advantages over Linux for a long time to come.

    Although Sun have made some mistakes with Solaris over the years I think they’ve always had the advantage of an OS and kernel which are technically better than Linux. Linux development (and much open source development in general) is somewhat unfocused so whilst there are more developers working on Linus than on Solaris their efforts are not always coordinated. So I think Sun’s focus means that they will maintain, and continue to extend that technical advantage. You already cited a couple of examples of where this is evident; Dtrace and ZFS.

  • http://www.redmonk.com/sogrady stephen o’grady

    Jaime: thx for the comments. to be clear, i’m not saying that package management is needed in every Solaris deployment, but to the extent that it’s competing for mindshare amongst developers in non-mission-critical, non-production scenarios, it loses on that score. i applaud – and have donated to – communities like BlastWave, but at the end of the day Linux has a better package management story currently than does Solaris. and that hurts, even if it doesn’t in production.

    James: no, of course not ;)

    Gyro: thx for the note; i’m actually aware of the Solaris patch facilities – i’ve used smspatch and Blastwave and Ben Rockwood had a good post on them previously. but none of the above – in my experience – are as flexible or as comprehensive as what i have available to me on Gentoo.

    to your point that the separate libraries maintained by the various Linux distros are a potential problem – and perhaps advantage for OpenSolaris – i agree. i’ve written that one up previously.

    Chris: very interesting feedback, would love to talk to you more about your experiences if you get a chance. drop a line to [email protected]. i can attest to the fact that the Solaris install process is less than ideal.

    erast: thx for pointing that out – sorry i omitted it.

    Philip: i don’t know that i agree. i’ve been saying for a while that when features that Solaris has truly make a difference – such as DTrace – begin to get customer traction and mindshare, you’ll see a focused response. i think we’ll see a concerted effort from the Linux community to close the gap in a couple of the areas. not all, perhaps, but Solaris will have to continue to innovate to stay ahead of where Linux falls short.

  • http://jroller.com/page/jaimec Jaime Cardoso

    Stephen, The roadmap is already established for Solaris 11 and, I’ve seen discussions about Solaris 12 so, Solaris will evolve.
    As for Linux, in the long run, I’m a lot more conservative, I think I’ll just have to wait and see how critical issues like the glib will be addressed.
    One thing is true, getting Solaris to the level Linux is today (in the things he is stronger) is a lot simpler than geting Linux to the level of Solaris (in the oposite things)

  • http://www.redmonk.com/sogrady stephen ogrady

    Jaime: to that point, what are you, as an active user of Sun technologies, seeing in Nevada and subsequent releases that’s functionally as innovative as, say, DTrace? i’m curious.

    as for getting Solaris to the Linux level versus the opposite, i’m not sure i agree. i think it’s much easier to copy – and improve on previously established features – than it is to build a community.

  • http://jroller.com/page/jaimec Jaime Cardoso

    Hey, comparing “stuff” with Dtrace is kind of hard. Dtrace is really ground breaking and I ever saw few things that compare but, I would put pretty much at the same level the Network performance increase that it’s beein worked out for Solaris 11 and the linear performance scalability increases (when “comodity” hardware has 64 “CPUs”, that scalability will be needed)
    As for the Linux – Solaris two way chase, for Linux to catch Solaris it will need to scale it’s Kernel up to way more than 100 CPUs, Solaris on the other hand, needs a new Package Management tool (just to use your example). That’s the diference I’m talking about.

  • iwan ‘e1′ rahabok

    Thanks for sharing your data points. I enjoyed reading non-Sun’s opinion/perception/etc. Customer’s perception, wrong or not, is my reality. Would love to read the complete report when you get there. A good time frame might be Q3 2006, because there are a couple of products we are launching (Solaris 10 Update 2 inclusive), which might propel the ‘acceptance’ further. Regards from Singapore, Iwan, IT Architect, Sun CSO, Asia South.

  • Pingback: People Over Process

  • Pingback: People Over Process

  • http://duckdown.blogspot.com James

    There are many enterprises who are considering moving back to Solaris from Linux. Maybe you could track down some of them and figure out the enterprise perspective.

    Likewise, the vast majority of the “appliance” marketplace used to run on Linux but also are starting to move towards Solaris.

  • Pingback: James Governor's MonkChips

  • http://phillipfayers.blogspot.com/ Phillip Fayers

    Philip: i don’t know that i agree. i’ve been saying for a while that when features that Solaris has truly make a difference – such as DTrace – begin to get customer traction and mindshare, you’ll see a focused response. i think we’ll see a concerted effort from the Linux community to close the gap in a couple of the areas. not all, perhaps, but Solaris will have to continue to innovate to stay ahead of where Linux falls short.

    Sun have an advantage over the Linux community, they have a more focused development agenda. That focus has led to the creation of Dtrace and ZFS. Now Linux also has its own innovations and these are, obviously, focused on the needs of the Linux users. So Linux has faster package installation because people are always installing packages on Linux, an area which Sun haven’t focused on so they are behind the curve.

    When I said that Solaris would stay ahead technically I suppose I ought to clarify that. Solaris will continue to have a kernel and supporting O/S infrastructure which is technically better for running enterprise and internet services.

  • DaveA

    You wrote: …credible advances in Solaris such as SMF

    Your kidding right? I have used UNIX over 25 years and I see no sense in editing a BIG XML file for a little config change. There ARE places XML is just no good for, UNIX config files are one of them. UNIX so far has avoided the Microsoft registry hell but it looks like Solaris has seen its peek.

    But the writing was on the wall in Solaris 9 with the naming of “directoryserver”. It tells tells me people are coding Solaris that don’t know UNIX, why not “dsadm”? Using long names and XML is a bastardization of the elegant simplicity of UNIX. Try writing a shell script to write much over verbose XML (and get it right the first time).

    If Solaris 11 goes further down the XML road, I will jump to Linux. Might anyway, already have it running.