back to article SPARC and Solaris will live until at least 2019

Oracle has quietly published a roadmap for its legacy Sun SPARC and Solaris platforms. Big Red's not offered a whole lot of detail, confining itself to the single slide below. The slide is available here as a PDF or as an embiggened image here. Oracle's SPARC and Solaris roadmap Oracle's SPARC and Solaris roadmap. Bigger …

COMMENTS

This topic is closed for new posts.
  1. Anonymous Coward
    Anonymous Coward

    The Oracle Walled Garden roadmap

    Meh!

    How much will they jack up the prices between now and 2019?

    IMHO, each and every price hike pisses off their dwindling customer base even more.

    but larry really needs a new mega yacht for delivery in 2020.

    1. Anonymous Coward
      Anonymous Coward

      Re: The Oracle Walled Garden roadmap

      I thought he only needed money to widen his moat.

    2. Anonymous Coward
      Anonymous Coward

      Re: The Oracle Walled Garden roadmap

      How much will they jack up the prices between now and 2019???

      Dude-time to wake up -its 2014! Oracles SPARC systems today are a fraction of the price/performance of the previous generations. A SPARC T5-4 system today outperforms Sun's previous HIGH END M9000-64. If you look for example at the TPC-H datawarehouse benchmark (at 3TB scale), the SPARC T5-4 achieved 409,721.80 QphH at $3.94 $/QphH while the M9000-64 achieved 386,478.30 QphH at $18.19 $/QphH just two years earlier.

      http://www.oracle.com/us/solutions/performance-scalability/sparc-t5-4-record-tcph-1956182.html

      That’s a 6% improvement in performance with 16 x fewer CPUs, 192 x fewer cores and almost 5x better price/performance, which includes Oracle SW licensing. Don't know any other architecture that has seen such a radical improvement in price/performance in last 5-years and certainly doesn't look like its in right direction for Larry considering the SPARC T5-4 requires 4x fewer licenses (that’s a $4.5M savings in Oracle DB licensing costs @ list price (192 fewer cores x .5 multiplier x $47.5K/license))!

      1. Michael Wojcik Silver badge

        Re: The Oracle Walled Garden roadmap

        I find it hard to muster any enthusiasm for SPARC these days, but have an upvote for citing some actual figures and doing the arithmetic. These interminable my-CPU-is-bigger arguments on the Reg are usually filled with completely fact-free posturing.

  2. Evil Auditor Silver badge

    They are still around?

    I have a soft spot for SPARC & Solaris. Solaris was my favourite OS, both for its architecture and GUI, and out of nostalgia I cannot get rid of my old SPARCstation. But seriously, I thought they were long gone.

  3. Anonymous Coward
    Anonymous Coward

    Better than I expected

    Just returned to Solaris (11) after a few years playing with smaller RH boxes, have to say I'm pretty impressed, on the T5 it flies and the built in virtualisation integrates very well, got 128 core VM's running 10,000+ threads and its hardly breaking sweat.

    Is it too late? I don't know - its a really good product and way more scalable and advanced than any Linux I've used, but they need to improve the pricing, marketing and most importantly customer service way beyond how its been since they swallowed Sun. Linux will always be the volume play of course.

    1. Alan W. Rateliff, II

      Re: Better than I expected

      Customer service: yup. Still can't get a support contract for my machines.

      @Evil Auditor: I have a few Sparc machines which I still have fun with. An SS20 with dual 200s for home play, a SS20 clone (Axil 320) with dual Ross 100 and 512MB RAM which is an actual production machine, and an IPXstation which runs like a champ but takes FOREVER to compile anything useful (and is also limited to Solaris 7.)

      Really fun stuff. I feel more comfortable with Solaris than other Unixes, for whatever reason.

  4. Anonymous Coward
    Anonymous Coward

    rip off

    Their pricing is an issue their support costs are getting to be a total joke. Hence we now have one single SPARC SOLARIS box which will be going in the next year or two. As a research lab we used to have loads, workstations and servers, lots of ultra 60's and 10's servers running NIS, firewall, backup, all gone. Workstations replaced by DELL Precision workstations and the server by DELL servers running various Linux installs. They're pretty much going the way of SGI (used to have lots of their kit too)

    1. Anonymous Coward
      Anonymous Coward

      Re: rip off

      lots of ultra 60's and 10's servers running NIS, firewall, backup, all gone.

      Doesn't sound like you're the target market now, Larry is concentrating on big systems that blow IBM out of the water. There's no money in the little stuff these days.

  5. sawatts

    Solaris-SPARC - a very capable platform

    25 (!) years using and developing on Solaris/SunOS systems, as well as Windows, Linux, and others; and I always considered Solaris-SPARC to be a rock-solid platform; must more capable than Windows for networking and process management (ie server activities) (vastly more capable!).

    Pressure to migrate software to the Windows monoculture - porting was like stepping back two decades.

    The annexing of Sun by Oracle has been a great disappointment; the first Solaris update after Oracle took over broke backwards compatibility with our software at the time - something which had never happened under Sun - at least in my experience.

  6. This post has been deleted by its author

    1. Anonymous Coward
      Anonymous Coward

      Re: Linux is the answer

      Let's face it, nobody is deploying *new* workloads on legacy operating systems such as Solaris, HP/UX, VMS, Windows Server, and whatever IBM is calling their mainframe OS now.

      Let's face it, that is simply nonsense. There is no Linux which can support the kind of massively multi-core/multi-threaded applications that a Solaris/SPARC (or equivalent IBM) type of system can. Linux is great for small systems, I run it myself, but most Linux users seem to think that a "high-end" system has a couple of multi-core CPUs, a fancy GPU, and a few tens of GB of RAM. Try to run Linux on a system with hundreds of multi-threaded cores, tens of TB of RAM, and a database that is hundreds of TB in size, and it will drown. The kernel simply can't cope, it doesn't have the sophistication.

      That, of course, means that Linux can be much faster than Solaris on a small system, there is less overhead in the kernel to deal with the fine-grained management that is required to handle tens of thousands of threads on hundreds of cores. As always it's horses for courses.

      As for balkanisation, the Linux distro situation is far worse than anything we saw with Unix in the 80s/90s. They may all run some version of the same kernel, but the rest, especially the management and UI side, is a fragmented mess.

      1. eswan

        Re: Linux is the answer

        >There is no Linux which can support the kind of massively multi-core/multi-threaded applications that a Solaris/SPARC (or equivalent IBM) type of system can

        Err, http://www.top500.org. Even the sparc powered Fujitsu K runs linux.

        #1 Tianhe-2, 3,120,000 cores, Kylin Linux

        #2 Titan, Cray Linux

        #3 Sequoia, Linux

        #4 K computer, Linux

        etc.

        1. Victor 2

          Re: Linux is the answer

          You clearly don't understand what "high-end" means if you cite HPC.

          1. Ramazan
            Paris Hilton

            Re: You clearly don't understand what "high-end" means if you cite HPC

            "High-end" deals with LPC (low performance computing) then?

            1. Victor 2
              FAIL

              Re: You clearly don't understand what "high-end" means if you cite HPC

              I have a dilemma... I'm not sure if I should humiliate you publicly for having no idea what a commercial workload is, and how high end and HPC differ from one another, or if you did that on your own stupidly enough. Paris is a good icon for you, as it's your best description, Ramazan.

      2. pwl

        Re: Linux is the answer

        The scalability discussion is no longer relevant.

        HP's "Kraken" server is a single-system-image machine with up to 240cores and 12TB of RAM (24TB once bigger memory modules are available). It's designed to run SAP HANA, and that's developed on and only available on SUSE Linux.

        http://www.v3.co.uk/v3-uk/news/2268742/hp-teams-up-with-sap-to-release-kraken-inmemory-big-data-system

        On the insane side, SGI's Ultraviolet systems can scale with a single system image of 2048 cores and 64TB RAM. Again, the biggest of these systems have been run with SUSE Linux.

        http://www.sgi.com/products/servers/uv/

        So, from a former Solaris OS ambassador: the scalability discussion is no longer relevant.

        1. Anonymous Coward
          Anonymous Coward

          Re: Linux is the answer

          Don't really agree as X86/Linux scalability/reliability hasn’t been proven with enterprise/commercial (non HPC) workloads at the highest non-clustered, single OS capacities/scalability levels. And of course, wheres the performance proof points???

          Your article on HP Kraken is almost a year old and still now word on when it will be a real product? Its certainly vaporware today. SGI's UV systems are optimized to run HPC workloads. SGI UV is non existent in the enterprise markets where DB, apps, java, etc are required. Wheres the benchmarks or public proof points otherwise?

          240-cores/system? Oracle has been shipping 16-core/CPU in SPARC T5-8 for over a year now with 128-cores, 1024-threads and 4TB (soon 8TB) support and SPARC M6-32 scales to 32 x CPUs, 384 x cores with a massive 3072 threads and 32TB RAM, all running Solaris, and any Oracle, IBM, SAP SW workload.

          Have you seen the world record SPARC M6-32 SAP SD 2-Tier result?

          https://blogs.oracle.com/BestPerf/entry/20140327_m6_32_sap_sd

          Or the SPARC T5-8 TPC-C World record for fastest non-clustered OLTP server?

          https://blogs.oracle.com/BestPerf/entry/20130326_sparc_t5_8_tpc

          Or the SPARC T5-4 worlds fastest TPC-H non-clustered datawarehouse @ 10TB?

          https://blogs.oracle.com/BestPerf/entry/20131125_t5_4_tpch_10000gb

          Or the SPARC T5-2 worlds fastest SPECjEnterprise2010 2-CPU Middleware server?

          https://blogs.oracle.com/BestPerf/entry/20140123_t5_2_specjenterprise2010

          Or the SPARC T5-8 worlds fastest SPECjEnterprise2010 Middleware server(s)?

          https://blogs.oracle.com/BestPerf/entry/20130926_jent_sparc_t5_8

          https://blogs.oracle.com/BestPerf/entry/20130326_sparc_t5_8_specjenterprise2010

          Don't see any SGI UV, HP Kraken or any x86/Linux (non-clustered) systems at the top of any commercial workload benchmark...

  7. Anonymous Coward
    Anonymous Coward

    Really?

    Why is this such a negative story? The T4 & T5 processors are really good and a huge improvement over the earlier T-series processors (which suffered from poor single thread performance). My company uses a lot of SPARC and I'm personally really looking forward to the next generation software on silicon features (hopefully they will release more details at Hotchips this year).

    I guess I just can't see why you would use such a negative tone. Oracle have released five new processors in the last four years (since they took over Sun) and are showing a roadmap for releasing a bunch more. Shouldn't us IT folks celebrate new technology? The Register would be a pretty boring read without new product releases wouldn't it!

  8. Victor 2
    Facepalm

    So, according to this article's title...

    Power processors will live until 2015 and Itanium is already dead?... Do you grasp the term "roadmap"?

    1. Anonymous Coward
      Anonymous Coward

      Re: So, according to this article's title...

      Atleast Oracle has the balls to show a 5-year public roadmap with actual dates and product milestones including performance gain estimates! You can bet that the competitors and analysts will all attack Oracle when they don't deliver to this very public roadmap, so I guess Oracle has their delivery reputation on the line.

      Neither IBM nor Intel is willing to stick its neck out...

      Looking at IBM's Roadmap, Slide 4 in link below, beyond Power8 is questionable with no dates, no performance expectations or new technologies. From what I can see on the slide, Power8 will take a year to be released, and that’s it? Doesn’t IBM have Power plans for next 5 years? Can it prove it to customers/press/analysts?

      https://www-950.ibm.com/events/wwe/grp/grp030.nsf/vLookupPDFs/Tour%20P8%20Charts/$file/Tour%20P8%20Charts.pdf

      And looking at Intels "DCG" Xeon roadmap on slide 15, the mission critical line, including Xeon E7-v2 series, which was just announced, is that all that’s on the "roadmap" going into 2015 and Future? Whats next? And what new technologies are planned in future? And why isn't there a tick/tock like on the entry level? Sure, we all believe in Intel delivering, but looking at the EX line, and Westemere-EX being 3 years old, wondering if Intel still focused on high-end x86?

      http://www.intel.com/content/dam/www/public/us/en/documents/roadmaps/public-roadmap-article.pdf

      1. Anonymous Coward
        Anonymous Coward

        Re: So, according to this article's title...

        And we shouldn't forget those who wish to be considered "enterprise suppliers" who also don't release product roadmaps, namely Microsoft and the open source community.

  9. Captain Server Pants

    Looks like tock then tick

    First a die shrink of existing T5/M6 but with new features and then a new core using same fab process. Pretty much what was expected.

  10. MemyselfandI

    Looks like I lost my bet - still not sure why

    After Oracle acquired Sun in 2010, my bet was that Larry would soon enough realize that HW development would best be left in the hands of a competent partner, like Fujitsu, and that within a few years he would send the majority of those well paid Sun engineers home, and focus future development efforts on OS, database, middleware, applications, and the like. After all, that is where Oracle's competence as a SW company really lies. When Fujitsu prominently showed their forward looking SPARC roadmap at Openworld 2013, and the new Fujitsu servers were introduced on Oracle's price list during 2013, I figured - any day now - and SPARC HW development at Oracle will be history.

    Well, I suppose I was wrong, and I'm not sure why. HW development is a very expensive business because developing new portfolios of enterprise servers is a major undertaking that consumes a large number of expensive resources. You have to be able to sell pretty significant volumes at outstanding margins just to make your money back let alone make a profit. Oracle being clearly very profit oriented, I figured this won't work out in the long run, particularly when one has choices to things differently. Thoughts anyone regarding big picture SPARC strategy, or is this something that has run its course and most no longer care?

    1. Anonymous Coward
      Anonymous Coward

      Re: Looks like I lost my bet - still not sure why

      Heres why MemyselfandI...

      Remember when Oracle acquired Sun, Oracle came out with an AD in WSJ?

      http://www.oracle.com/us/features/suncustomers.html

      Theres a great quote in there "We're in it to win it. IBM, we're looking forward to competing with you in the hardware business"-Larry Ellison, and since then, roughly 20% of Oracle's annual R&D has been focused on delivering industry leading hardware, to beat IBM, and of course all the others. So why invest in HW when IBM, HP, Dell and even Intel have been suffering so badly, with eroding margins?

      Well, the way I look at it, is basically what Warren Buffet said a few years ago "“In business, I look for economic castles protected by unbreachable ‘moats’.” where he then further said "Make the moat your first priority. Protect the business.  Then the profits will come”.

      Oracle's "Moat" is its IP and according to Oracle's Fact statement, they have over 400,000 (Mostly SW) customers, where over 300,000 are running Oracle DB. Its clear that before the Sun acquisition, Oracle was atleast 95% focused on SW, but I believe they and Larry of course realized, the only way to protect that SW base, or "moat" in Buffets words, and to protect it from competitors, and for Oracle to continue growing that business would be by defending it from other SW vendors, many of them who sell HW.

      What better way to do that then by developing hardware systems, like engineered systems and especially SPARC systems, where Oracle can engineer/optimize/tune this hardware, primarily for its SW, so there would be no other platform that would run that SW as well, nor get the same ROI on that SW, as it would on Oracle HW. Sure choice is still important, and supporting all major platforms, but ultimately, if you run Oracle SW, the only way to get the best performance. best integration, optimization, etc, is running on Oracle HW. And when you run Oracle SW on Oracle HW, you get all the OPEX savings as well-lower maintenance headaches, fewer patches, less tuning needed, integration challenges, support issues, etc.

      So if you look at Oracle latest financial reports, you'll see that since the Sun acquisition, Oracle has continued to gain not only in Revenue, but also in margins and profit? Why?How? Because its systems, although slightly more expensive than all those commodity players out there, show a significant price/performance and TCO advantage when running Oracle SW that is optimized/integrated/engineered better for Oracle specific hardware, and so in the end, becomes lower cost. Look at the latest Gartner or IDC reports and you'll see that Oracle is leading in "Integrated Systems" with roughly 50% marketshare, and that Oracle recently announced a 10% increase in Hardware sales revenue while IBM, for example, has been declaring double-digit losses qtr after qtr. Even with SPARC, according to IDC, grew 6%. Heres an article with the details : http://www.computerweekly.com/news/2240216360/Oracle-pushes-Exadata-servers-aggressively-against-IBM-PureSystems.

      And finally, we see Fujitsu also heavily investing in SPARC and aligning with Oracle. Why, because customers like choice and options, and having two major investors in SPARC means goodness for ISVs developing for the platform. Kind of like AMD and Intel. Difference is, Oracle and Fujitsu are 100% focused on the commercial/enterprise.

This topic is closed for new posts.

Other stories you might like