Earlier this year, Oracle released its Sparc T5 processors and a range of server configurations to house them. In doing so, Oracle demonstrated it acquired not only Sun Microsystems' strengths, but also some of its flaws: the Sparc processor promptly set all sorts of performance records and is clearly backed by a strong engineering discipline and an excellent software stack in Solaris and the Oracle suite, but the market doesn't seem to be buying, and revenue continues to slide.
The T5 announcement was significant in several ways. It pulled the Sparc roadmap back on track (ahead of schedule, even), and lay the groundwork for the Oracle optimisations expected in the T6 (due in 2014). With Sparc development now firmly under the control of Oracle, the company can ensure the processor, although aimed at general-purpose data centre computing, is optimised for database and Java operations to give a further performance edge when operating in the company's core environments.
The T6 (probably the first processor developed from scratch by an Oracle-owned Sparc team - processor design is a many-year process) is expected to have much more database capability baked into the silicon, and promises another major performance boost even before that's taken into account.
Sparc, Oracle therefore insists, is far from dead. It is being actively developed, strongly promoted, and partner Fujitsu is still very much on board (eyebrows were raised at the Japanese firm's apparent exclusion at the T5 launch, but the companies are still actively cooperating).
Top of the charts
On paper, there's no reason why the Sparc family shouldn't do well. Its performance (and, importantly, price/performance) set the bar high, pulling back its rivals in a way many thought unlikely. The T5 launched with 17 performance records, though chief rival IBM was quick to disparage some of them.
Oracle, under the leadership of CEO Larry Ellison, is never one to back down from a war of words, so the companies have been exchanging barbs ever since, played out in the media and numerous insider blogs.
"Benchmarketing" is a popular tool in high-end computing, but even if some of the numbers are cherry-picked, there's little doubt that the Sparc T5 is a good processor. And the advanced features in its server housings, such as high-speed interconnect and fault tolerance, coupled with the well-regarded Solaris OS, should give it, if not a head start over its rivals, at least a strongly competitive offering.
But does it matter? One of the reasons the companies are squabbling over numbers is that they are fighting over slices of a shrinking pie. The RISC market isn't big. Oracle and partner Fujitsu are duking it out with IBM, HP, and Bull. The products are high-end computing centres, running line of business applications and server consolidation projects.
But across the board, the server market is struggling, with shipments and revenues under pressure, especially from the cloud. Server revenue ticked upwards earlier this year, but that was largely thanks to margin boosts in IBM's mainframe business - probably not a sign of a market-wide recovery. But RISC-based revenue has taken an absolute beating. In EMEA, server revenues were poor, but RISC revenues dropped an astonishing 31.7% year-on-year in Q4 2012, Gartner reported.
Commodity hardware has seen the server world move steadily to Intel, and the cloud (itself frequently Intel-based behind the scenes) is compounding the pain. The server space is no longer about the choice of platform, but also about ideology. Are you building into, or for, the cloud? Are you willing to bet on a single vendor's stack and partner ecosystem? Low-cost/high-density clusters, or high-end behemoths?
Playing it safe
Oracle, like most manufacturers, is hedging its bets with multiple options, but the database giant is playing in almost every camp all at once. Most server vendors have multiple product lines - the RISC vendors all offer Intel products too, and all offer integrated solutions. Dell, for example, has been aggressively acquiring companies for years to build out its stack. Oracle has the Intel-based range of high-performance integrated solutions, including the Exadata and Exalogic appliances, alongside the SPARC kit for multipurpose RISC computing, and of course, it has the whole software stack which you can run on any hardware you like. And it is also pushing hard into cloud computing, with its own platforms tuned for cloud providers (and enterprise customers deploying private clouds), as well as public cloud services.
But no matter how good the new Sparc chips are, there's no question that the RISC target audience is shrinking, defecting to Intel (which has Ivy Bridge Xeon upgrades of its own in the pipeline this year) and to cloud services. Despite that, there are still plenty of existing Sparc customers, and analysts say enough of them are upgrading their infrastructure to make it worth Oracle's while. Oracle will be hoping for more: that the T5 and M5 systems can claw back a larger slice of the pie. But the uncomfortable fact will remain that the pie as a whole is still shrinking.
If a reversal of fortunes does come, it won't be a moment too soon. Oracle is currently losing ground faster than its peers, and not just in RISC. Its hardware revenue and market share is under pressure across the board. CEO Ellison has promised that 2013 would be the year the hardware business turned around - the spotlight will be on that claim during earnings reports. In the company's last report, hardware sales had dropped 23%.
Like most of its competitors, Oracle is betting heavily on its stack - building engineered systems designed to enfold customers within an ecosystem. Oracle's walled garden is more heavily fortified than most, with the company intent on offering every piece of the system from services to bare metal. But for all its engineering excellence, it's just not making the sales.
Oracle as a company is in a healthy position - it has strongly entrenched positions in the database world, and has a strong application suite and services business. Its stock price dipped at its last earnings call, but bounced back to its healthy multiyear trend. Even if, worst case, its hardware business were to tank terminally, Oracle would still be a good, profitable business.
Server vendors of all flavours are under enormous pressure, and looking to create more resilient businesses by expanding their hardware solutions amid a turbulent phase in high-end application hosting. It remains to be seen whether Ellison's fixation on dominating every slice of the stack will translate into a revival for the hardware business, but Oracle has its fingers in enough other pies to benefit whichever way the market moves.
But Sparc? That future is uncertain. If you're a Sparc shop, are you upgrading? Investing? Migrating? Post a comment and tell us which way you're moving.
Share