Oracle says it's finally done upgrading SPARC servers since Sun acquisition

Oracle says it's finally done upgrading SPARC servers since Sun acquisition

Summary: Larry Ellison revealed that Oracle is planning to start moving many features of its software and database portfolio onto silicon and hardware.

CEO Larry Ellison at Oracle OpenWorld 2012 last October.

REDWOOD SHORES, CALIF. -- Oracle has unveiled its latest generation of SPARC servers for mission-critical applications, boasted to be running on the world's fastest microprocessor.

See also: Oracle's Q3 falls short, revenue misses mark; Hardware systems tank again | Oracle's Q3 miss: Canary in enterprise software licensing coal mine?

Designed with cloud deployments in mind, the new midrange T5 and higher-end M5 servers have been built to enable "near linear scalability" from one to 32 sockets with a single common core, operating system, and common set of systems management and virtualization tools.

John Fowler, executive vice president of hardware development at Oracle, said plainly that the these servers are meant to take on "any workload" as well as be deployed to work together, mixed and matched across an enterprise.

CEO Larry Ellison offered a few hints about the new servers during last week's quarterly conference call with investors.

During that call, Ellison described the SPARC T5 as "the fastest processor in the world" for databases and Java middleware.

He added that Oracle's new T5 servers have up to eight processors, which are touted to be more than twice as fast as their T4 predecessors.

Even more importantly, Ellison stressed, the new M5 server, which has up to 32 processors, runs the Oracle database over 10 times faster than the similarly priced M9000 server it is replacing.

With the launch of the M5 server, Oracle is asserting that it has finished upgrading every server in the SPARC product line dating from the time it acquired Sun in 2010.

During the official unveiling on Tuesday at Oracle's headquarters, Ellison remarked that a lot of people thought the SPARC series was lagging at the time of the Sun acquisition in 2010.

Ellison boasted that Oracle took it, caught up, and surpassed the competition.

For example, the hardware giant asserted that the T5-8 is the fastest single server for Oracle Middleware with 12 times the cost performance advantage over the IBM Power 780 as well as seven times the price advantage over the same IBM configuration.

Looking forward, Ellison described that Oracle is planning to start moving many features of its software and database portfolio onto silicon and hardware.

In other words, Oracle is going to put database query and Java accelerators onto the chips.

"Software in silicon is redefining enterprise computing," Ellison summed up, also quipping that "modern hardware is 90 percent software."

Oracle's roadmap -- at least through 2014 -- includes upgrading the execution codes to free up cores so that other codes can execute, improving performance and efficiency overall.

Topics: Hardware, Data Centers, Data Management, Oracle, Processors

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • So thus dies Oracle...

    I remember other manufacturers that went that route - Teradata as I recall started with custom hardware support for databases...

    Now they use Intel.

    The problem both have is that hardware isn't static - as soon as you stop developing hardware to focus on putting the database in hardware, you loose. General purpose hardware becomes better, and handles the same data... but much cheaper.
    • I don't think they are going to stop using intel

      For the sun x86 servers. They're just updating the SPARC platform. They clearly see the main competition for this as IBM's power7 based offering. Going intel here probably wouldn't be the way to go.

      I wish them luck, I've always had a soft spot for the old sun and SPARC... Plus it's be nice to see a t500 top 10 that isn't just "blue gene" all over the place. That said... The current king using gpu's ... Very interesting
      • The Power 7 is general purpose.

        And is developed for general use - not dedicated database processing.

        Now that Oracle is focusing on database throughput, it will not be doing general purpose CPUs except as a sideline.

        It also traps the hardware... General purpose hardware will do things a bit differently, and designed for more flexibility.

        The problem Intel has is I/O bandwidth - and memory bandwidth to CPU. It is a horrible design.

        The Power line has excellent I/O capability... but needs more CPU/memory bandwidth.

        The last Sparc systems had a memory bandwidth problem.. and cache memory additions only make it worse (slows down context switching).

        Personally, I think it is time for a new architecture - something closer to the old data flow machines that might allow processors to be pushed into memory - then a 4K memory page might include 1K of application processors...
        • I think that's where cray was going

          Implementing gpu's into the mix with the athlons. To be honest gpu derived chips and arm are the only new architectures I see entering the fray. I can't see IBM giving up on their power architectures anytime soon, and oracle clearly intend to keep SPARC alive. As to the wisdom of that, I'm not so sure. I was reading before about SPARC haemorrhaging cash for oracle, but I can't find the article now.
    • Right about Teradata

      I worked for Teradata back around 1990 when they had custom hardware. It was revolutionary for its time, but they couldn't generate enough revenue to support continued R&D. To get money, they licensed all their intellectual property for a few million to NCR, and it was gone within seconds to pay the bills. A few months later NCR took over the company in a friendly buyout (it was spun off into an independent company in 2007).

      Even though Oracle is a much bigger company, it faces the same problems.
      • Size of the company matters

        A large company can make big transitions easier than a small company. Apple has switched processor families 3 times, has completely changed their OS.

        Putting well established behavior in hardware really improves performance.
    • The key is to put the right functionality at the hardware level

      Many companies do this. Ours certainly has. The benefit is huge both in terms of speed and power consumption. Even GP processors are using accelerators (AES encryption/decryption). Long ago it was floating point units.

      The key is to only accelerate well establishded primitives.

      Regarding Terradata. That isn't a good comparison. Oracle is much larger with a huge customer base. They can make such transitions successfully.
  • Too late...

    Sun has lots massive amounts of marketshare over the last two decades. At this point, it is just a niche player. From the current discussions with the Oracle team, the hardware is viewed clearly as another way to sell the Oracle DB. Here is the rub: Oracle DB is very high maintenance. The high maintenance means that the DB is deployed less frequently than other DBs that are available. Don't get me wrong, for the Oracle DB licenses in house, it is a long and hard decision to stop paying maintenance. However, all new DB requiring apps get a hard look at alternatives first. Oracle knows it, which is why increasingly its acquired products run on Oracle DB exclusively and the DB license is not included in the purchase of the app. A good example is anyone who ran Sun IM/AM portfolio and say used MySQL. Oracle killed that product and now one has to use Oracle IM/AM portfolio and wait for only runs on Oracle DB that is extra charge. I have no doubt others will see what is the obvious thing -> get Oracle out of your data center.