Reflections on a Flickering SPARC

By Charles King, Pund-IT, Inc.  January 25, 2017

A report last Friday in the San Jose Mercury News that Oracle was laying off 450 workers in its hardware division suggests that the proprietary silicon experiment the company began with its 2010 acquisition of Sun Microsoft is nearing the end.

It’s sensible from a financial point of view, especially for a company like Oracle that is demanding when it comes to business unit performance. In its most recent quarter (Q2 FY2017) Oracle reported that sales of hardware products (servers, etc.) were down -13% to $497M for the quarter, and down -16% in the previous six months to $959M. The company has also suffered double digit sales declines during the past five quarters.

Additionally, Oracle’s server business has long been absent from the upper “Top 5” reaches of the server market, and thus relegated to the “Others” category in market sizing studies by IDC and Gartner. You could say that its faltering results suggest that Oracle either didn’t deliver on or wasn’t especially serious about its promises to Sun hardware customers.

Considering the strategy espoused by Oracle executives—focusing mainly on engineered/integrated systems and database appliances—the latter interpretation is closer to being correct. High-end solutions certainly have their place at Oracle, especially in applications where optimizing performance of the company’s core database solutions is concerned.

But with sales of traditional Unix-based systems, including Oracle’s SPARC/Solaris servers, under continuous pressure, the company needed and yet failed to do considerably more.

POWER comparisons

Like what? Consider what IBM has done with its own POWER chip architecture, and the Power Systems they inhabit. In 2010, the same year Sun went on the block, IBM delivered its POWER7 processors, and began rolling out a full complement of new solutions. Next gen POWER8 arrived in 2013, and POWER9 was introduced at the HotChips conference last August and is reportedly scheduled for production this year.

As a result, IBM has achieved a dominant position in traditional enterprise-scale system sales, but more important are the extraordinary efforts the company has made to broaden POWER’s usage. In August 2013, IBM announced that it was open sourcing its POWER chip architecture, an effort supported by the OpenPOWER Consortium (later changed to the OpenPOWER Foundation, an organization it founded with Google, NVIDIA, Tyan and Mellanox with the goal of utilizing POWER in new products and innovative use cases. Today, OpenPOWER has over 250 active members.

Beginning in 2012, IBM also began delivering PowerLinux servers pre-integrated and optimized for Linux workloads. The new systems offer double the compute threads and higher utilization than equivalent Intel-based systems, thus supporting enhanced performance for Linux-based business and technical applications. PowerLinux solutions also leverage other homegrown IBM technologies, including PowerVM, a virtualization solution derived from the company’s z Systems mainframe platform.

Isn’t comparing what IBM has done with POWER to Oracle SPARC a fool’s errand? Perhaps, unless one remembers that IBM was reportedly the first vendor in line to acquire Sun in 2010. According to rumors at the time, the deal was all but done until last minute changes to the agreement demanded by Sun executives caused IBM to step back.

Oracle stepping in with a counter offer was natural enough. A large majority of Sun customers ran Oracle databases, and the company saw huge commercial opportunities in Sun’s software portfolio. Many of those plans—suing Google for its use of Java, in particular—haven’t worked out quite as planned. Plus, despite employing experienced, high-profile execs, like Mark Hurd and David Donatelli Oracle’s hardware business is no one’s idea of a rousing success.

Overall, it’s hard not to think that Sun’s customers would have been better served if IBM’s acquisition plan had succeeded.

Final analysis

It would be a mistake to lay all the blame for SPARC’s decline on Oracle. The fact is that sales of Sun’s UltraSPARC-based servers were in trouble long before Oracle came along. That was partly due to the company failing to fully bounce back from the bust before getting whacked by the financial meltdown following the subprime mortgage crisis.

Sun also tried to resist the Intel-based religious conversion that gripped most the IT marketplace. Ineffective nods to Intel’s rising profile among enterprise buyers—paying over $1B for Cobalt Systems was a bad deal of legendary proportions—did nothing to get Sun on track. The company couldn’t even gain an advantage from its x86 version of Solaris—the only proprietary Unix variant ported for the Intel architecture.

To its credit, Oracle stated last June that it will deliver services based on its newest SPARC7 processors via its cloud platform, a point that should comfort customers who are diehard SPARC/Solaris users. But after the company released its most recent SPARC/Solaris roadmap, numerous reporters noted that the upcoming versions of both the processor and OS had been rebranded “Next” suggesting that neither would be a generational upgrade. If that is indeed the case, expect the flickering light cast by SPARC and Solaris to continue to dim.

As Sun eventually discovered, ignoring or avoiding technologies that your core customers want is the equivalent of slow motion suicide for IT vendors. But failing to continue developing established technologies and adapting them for new applications, changing use cases and evolving business requirements also results in a leisurely dangle at the end of a rope.

That, as Oracle, its hardware division employees and system customers have discovered, is an especially painful lesson to learn.

© 2017 Pund-IT, Inc. All rights reserved.