CNET también está disponible en español.

Ir a español

Don't show this again

Tech Industry

Commentary: Oracle faces challenges with 9i

Despite adding ambitious capabilities to 9i, Oracle still faces challenges with its first new version of its database software in two years.

    Despite adding ambitious capabilities to 9i, Oracle still faces challenges with its first new version of its database software in two years.

    9i provides new functions in areas such as OLTP and decision support and brings Oracle even with DB2 and SQL Server in OLAP performance. However, adding capabilities such as cube building in the DBMS and Real Application Clusters (RAC) are keys to this new release--and the most significant technical challenge.

    If this extension of Oracle Parallel

    See news story:
    Oracle takes on foes with new price plan
    Server works, it will be a significant advancement, providing the capability to expand a database with new logical nodes in a shared environment on any platform that Oracle supports. Logical clustering is difficult in an OLTP environment, and Oracle is the only vendor to even attempt it on a platform it does not totally control.

    However, that "if" is significant, given the short beta period Oracle gave 9i to meet Larry Ellison's promise to deliver the new database version in the first half of this year. We recommend that users wait 10 to 12 months before moving production databases to 9i, just as we would for any major DBMS migration. But Oracle should get some credit for attempting this feat. In effect, it is demonstrating its leadership of the database industry just by making the attempt.

    However, we do not expect RAC to work immediately without a hiccup in production systems. We expect that Oracle will need to bring out intermediate upgrades, and that it may not work outside specialized applications on any platform until clustered file systems are available for the four major platforms. This should be available from Veritas around the same time as the next scheduled major upgrade of 9i, which is due in March 2002.

    A key question is when Microsoft will come out with a cluster file system. Oracle wants to include cluster file system functionality in RAC so that it will work on any platform, but we believe it will take a significant amount of time to achieve that.

    Combining this with the introduction of Intel's new 64-bit Itanium processor family creates the possibility of commodity Unix systems. Hewlett-Packard and IBM have promised to port HP-UX and AIX to Itanium. But, significantly, Sun has no plans to move the SPARC environment.

    If Oracle does succeed with RAC, it will be a major breakthrough in database capabilities. Although Oracle talks about it mainly as an advance for OLTP, we see its most valuable use in high-availability situations requiring a master/slave environment. We also expect early adoption for analytical applications, which currently are some of few applications that use two nodes under Oracle Parallel Server. But it does not deliver the performance that IBM's and NCR's shared-nothing architectures deliver.

    We believe Oracle has a good chance of making RAC work. From an architectural standpoint, it looks sound, and when it does work it will be a huge leap. For example, users that currently have a high-availability application on an eight-way system typically have a second eight-way system standing by in slave mode. If these users want to move to a 16-way system, with RAC they will no longer need a second 16-way just to stand by in slave mode.

    Users should not plan to move production databases to 9i for at least six to eight months, giving Oracle a chance to iron out the inevitable wrinkles associated with a release this significant. Risk-weary users should wait until the second major version appears in March 2002. Those with advanced needs demanding leading-edge technology (such as high-availability master/slave databases) should use the intervening period for testing.

    In general, however, we expect 9i to have significant issues. Sixty days is too short for a general beta test of major enterprise software, particularly because it would take the testers several weeks just to get their hands on it and get around to installing it. Users should regard the next 10 months as the real 9i beta test.

    META Group analysts Mark Shainman, Charles Garry, Val Sribar, Dale Kutnick, and David Cearley contributed to this article.

    Visit Metagroup.com for more analysis of key IT and e-business issues.

    Entire contents, Copyright © 2001 Meta Group, Inc. All rights reserved.