eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
As Oracle Corp. girds for the upcoming release of its Oracle 10G database next month, the company is cranking up the G-for-grid part of the message for this major update.
According to beta users, the update will feature easier-to-configure grid capabilities. But a big part of Oracles RAC (Real Application Clusters) message for the past year has been about Linux, which may shave the cost of the platform by up to five-sixths, according to Wim Coekaerts, director of Linux engineering for Oracle, of Redwood Shores, Calif.
One question remains: Is Linux really enterprise-database-ready? Some experts say no, pointing to the fact that the operating system was designed as a version of Unix that would run on a 486 PC, not as the platform for an enterprise-level database, never mind a whole grid of enterprise databases.
“It was from the get-go not designed as a database platform,” said Andrew Binstock, an analyst at Pacific Data Works LLC, in San Carlos, Calif. “There are elements that still need further development to become a truly enterprise-scale database platform.”
A key component that needs further development is the threading model used by the system kernel, Binstock said. Multithreading is crucial to running an enterprise database. Without it, a database would have to handle every data query as it comes in, keeping all other queries waiting in queue, since the system would be incapable of parallel processing.
Threading support came to the Linux kernel only within the past few years, and even now there are kinks that need to be worked out. For example, even though the upcoming Linux kernel, 2.6, features additional support for threading interfaces—also known as Pthreads—the support is not yet complete, Binstock said.
Discuss this in the eWeek forum.
Next page: Whos using Linux now for enterprise databases?
Whos Using Linux Now
?”>
But some users are in fact running enterprise databases on Linux. Patrick McGovern, director of SourceForge.net, is one of them. SourceForge.net is the largest open-source development Web site, experiencing exponential growth since its inception in 1999 and now hosting in excess of 66,000 projects.
Because of scalability issues, SourceForge.net has made a commitment to move off PostgreSQL and onto IBM DB2 databases running on Linux. The Linux kernels supposed drawbacks havent hampered this busy installation, which is adding some 700 new registered users daily, said McGovern, in Fremont, Calif.
“I think its gotten a lot better in the last couple of years,” he said. “Red Hat [Inc.]s Advanced Server adds a lot of additional kernel tweaks and stuff to give it higher performance to run a database, with things like [additional support for] threading and memory allocation.”
Its not a good idea to put an enterprise database on Linux without Advanced Server, experts say. “I know lots of people who run on Linux,” said Ian Abramson, chief technology officer of Ian Abramson Systems Inc., in Toronto. “They run fine. But … if you start going into the unknown or into the new or newly developed, youll run into issues. With Advanced Server, you minimize that.”
Others disagree. Craig Read, president of the Toronto Oracle Users Group and IT director at MTrilogix Inc., in Toronto, said that even with Advanced Server, he wouldnt recommend to his clients the use of Linux for a distributed and robust environment where backup and replication are essential. Weak points include lack of available tools, ease of use and ease of installation, Read said. “I dont think Linux is there yet, nor is the kernel that robust to do really heavy grid-type computing or to be fail-safe. Not yet,” he said.
Brian Stevens, vice president for operating system development at Red Hat, in Westford, Mass., said many issues with the Linux kernels enterprise-database-handling capabilities have already been fixed or will be worked out in the upcoming kernel, expected in the next year. Specifically, 2.6 will have much larger memory support, enabling databases to cache much greater amounts of memory and hence run much faster, as well as enhanced I/O scalability, which should eliminate contention issues that lead to bottlenecks.
Discuss this in the eWeek forum.