Oracle continues to release information in dribs and drabs about its Sun Microsystems assets, which has made its...
user base uneasy. Recently it announced plans to release Solaris 11 next year, plus a roadmap for the Sparc chipset through 2015. But the Redwood Shores, Calif.-based Oracle Corp. left just as many questions about its Sun integration plans unanswered, leaving customers to wonder -- and wait.
In a webcast this week, John Fowler, Oracle's executive vice president, made no mention of OpenSolaris, Oracle's OEM relationship with Fujitsu, its plans for new server products, and whether it will introduce an Exadata appliance based on Sparc or Solaris.
"We know a little bit more than we did a couple of days ago," said Jean Bozman, a research vice president at IDC. "What we don't know is the productization." Bozman said she expects Oracle to leak more information around the time of Oracle's confab, OpenWorld, in September.
Oracle Solaris integration
For the forthcoming Solaris 11, the name of the game is integration with the Oracle application stack. Solaris 11 will lay the foundation for running high-performance, resource-intensive applications such as Oracle database with support for thousands of threads and terabytes of memory, as well as enhancements to its networking, virtualization, scalability, packaging and file system capabilities, Fowler said.
"Now that Oracle has acquired Sun, they can do things that they couldn't have done to Solaris before," said IDC's Bozman. She said Oracle can improve performance, flash memory support, the ZFS file system and hybrid systems. "You can do different things now that this technology is all in one basket."
Part of Solaris' scalability will come from enhancements to the Sparc chipset, whose application performance Oracle is committed to "at least doubling" every 18 months, Fowler said.
Fowler cited a Sparc processor that today has 32 cores, can support 512 threads, 4 TB of main memory, 128 logical domains, and deliver database transaction processing manager (TPM) and Java ops per second of 3 million and 5,000, respectively. By 2015, the processor family will support 128 cores, 16,384 threads, 64 TB of memory, 256 logical domains and deliver 120 database TPM and 50,000 Java ops.
"This is more roadmap than we had before," said IDC's Bozman. "We knew since Jan. 27 that they would invest in Sparc, but we didn't know what that meant." Now we know that at least when it comes to this particular Sparc chip, the UltraSparc with chip multithreading (CMT),"Oracle plans to evolve that design and make it more all-purpose with more cores and more threads."
Whither Fujitsu, new systems?
Meanwhile, the fate of Sparc64 chips and the M-series systems based on them is also unknown and will remain so until Oracle clarifies its relationship with Fujitsu, which manufactures and designs the Sparc64 chips and systems.
Large Sun shops have eagerly awaited that news, said Phillip Jaenke, an independent consultant who specializes in Unix and virtualization. For instance, the midrange M-series "is getting a little bit long in the tooth," Jaenke said, and Oracle has yet to propose a credible update.
"The dogs are at the door," agreed Mike Fitzgerald, the managing director at Eagle Investment Systems, a software vendor that provides hosted and on-premise software for the financial services industry. Eagle runs much of its database-oriented code on Sun, "but the Web tier is completely gone: There's no reason to run that on an expensive Sun box," he said, and large swaths of the application now run on Linux on Hewlett-Packard Co.'s hardware, he said.
"We're fully committed to the platform, but the biggest threat is the TCO of Solaris and Sparc compared with Linux running on commodity systems," Fitzgerald said. "You're seeing the gap in performance shrink, but the gap in the price points remains vast."
Small appetite for Solaris 11
Many Sun shops are content on Solaris 10 and, thus, haven't eagerly anticipated Solaris 11. Eagle Investments' Fitzgerald said he would upgrade to stay current, "but we're fine with the existing toolset in Solaris 10."
Then there's the question of what -- if anything -- Solaris 11 includes that is of interest outside the Oracle environment, said Jaenke.
Furthermore, the uncertainty surrounding the open source OpenSolaris has made members of the community nervous, said Jaenke. Like Fedora is to Red Hat, OpenSolaris is the "playground" in which developers test new features that may eventually make it into commercial Solaris releases, Jaenke said. But Oracle has not provided updated components to OpenSolaris since February, nor attended OpenSolaris governing board meetings, which has put the community on edge.
"There's been radio silence from Oracle; they've been invited to things and don't even show up," Jaenke said. "Since then, a lot of folks have gotten nervous and stopped contributing."
Making matters worse is the continual exodus of the Solaris brain trust. Bryan Cantrill, the inventor of the DTrace feature in Solaris 10, for example, left Oracle for cloud management startup Joyent. This exodus of knowledgeable personnel prompts observers like Jaenke wonder what value Solaris 11 can provide.
More to the point, Oracle could allay fears by laying off its hard-nosed tactics surrounding support and maintenance and by listening to customers.
"There seem to be two tracks within the organization," said Eagle Investments' Fitzgerald. "There are the people on the front line in sales and service that are determined to maintain quality of service to the customer," he said, "but there's a disconnect between front line and management, and it's unclear whether management will listen."