Oracle and storage IOs, explanations and experience at CERN

Abstract

The Oracle database system is used extensively in the High Energy Physics community. Critical to the efficient running of these databases is the storage subsystem, and over the years Oracle has introduced new ways to access and manage this storage, e.g. ASM (Oracle database version 10.1), Direct NFS (Oracle database version 11.1), and Exadata (Oracle database version 11.1). This paper presents CERN's experience over the past few years with the different storage access and management features, and gives a comparison of each functionality. Also compared are the different solutions used at CERN, and the Tier 1 sites for storing Oracle databases. 2. Oracle and the IO subsystem By nature, a database system is dependent on the Input Output storage system. The Input Output (IO) storage system is often called "disk subsystem", with the flash Solid State Drive (SSD) devices coming increasingly as replacement for the rotating drives, the IO subsystem is a more adequate term. The Oracle database system makes use of the different types of IO subsystems available, presented to the host, or hosts in case of clustering. In the past years, the operating system vendors or developers have added features to their system that have enabled the Oracle database to make better use of the computing resources and provide faster response time. This is for example the case of Asynchronous and direct IO features. A description of the major types of IO subsystems as used by the Oracle database is given is the next section.