Survey: Hardware usually causes Oracle database woes

22.06.2006
Faulty hardware, not hackers, caused most of the unplanned downtime experienced by Oracle Corp. databases in the past year, according to the results of a recent survey by the Independent Oracle Users Group (IOUG).

Despite the growing need for round-the-clock availability, more than seven out of 10 Oracle databases were down for more than an hour in the past year, according to the IOUG's April survey, which included 265 members -- most of them database administrators, developers or managers.

It wasn't natural disasters, power outages or security breaches that were behind most incidents, according to the survey. Security incidents led to only 1 percent of unplanned database outages. Power outages and natural disasters accounted for just 7 percent and 4 percent of unplanned database outages, respectively.

Hardware problems were behind almost half of the outages. Problems with the network, the application stack or the database itself contributed to about a third of the unplanned incidents, which could have more than one cause. Those experiencing both hardware and network problems were most likely to report that their databases were down for more than 15 hours.

Performance slowdowns were also common: 40 percent of respondents reported at least one slowdown every two months.

A full 80 percent said round-the-clock reliability was a necessity for at least some of their company databases. Moreover, 43 percent were governed by service-level agreements that required less than 24 hours of total downtime, including planned and unplanned, per year.

Even so, the use of Oracle high-availability solutions was not high: 31 percent used Oracle Real Applications Clustering, available in Version 9i of Oracle Database. Just 13 percent used Oracle Grid, available in Oracle Grid.

While nearly eight out of 10 respondents had either tape or disk backup, only 43 percent had formal disaster recovery or business continuity plans for their databases.

Planned downtime, such as when replacing a server or installing an operating system, database patches or upgrades, was also common. A full 39 percent reported losing 10 hours or more as a result of planned outages.

"The greatest challenge is to keep up with all the Oracle bugs and patch fixes," wrote one database administrator at a large government agency in response to the survey. "Even the current version of [Oracle] 10gR2 that we are on has a large number of bugs and issues. Keeping on top of these is a big challenge, considering that we have over 100 databases."