Preface

Improving Availability in a Non-Clustered Environment

The industry direction in improving system availability is the refining of clustered solutions. The fact that the industry is focusing on clustering as a solution to improve availability (moreso than a solution to scale performance) suggests that availability has not yet reached an ideal state. Clustered solutions utilizing ORACLE Parallel Server for NT do significantly increase availability. However, these solutions can be expensive in terms of purchase and maintainence costs, as well as in administration effort required. There are practical techniques that can be used to improve the availability of non-clustered ORACLE instances, the details of which are presented in this section.

It appears to be generally accepted that an ORACLE instance on NT running for 6 months (plus or minus 3 months) without an instance crash or O/S blue-screen is percieved as "acceptable availability", although the availability requirements will be specific to each company, application, etc. By the same token, several instance crashes per month or week is generally viewed as unacceptable availability. If your installation is in the unacceptable range but wishes to be in the acceptable range, then implementing the steps outlined in this section will tremendously assist in that effort.