Abstract
In this hint and tip, we describe the main benefits of an Oracle9i RAC implementation over GPFS.
For related information about this topic, refer to the following IBM Redbooks publication:
Deploying Oracle9i RAC on eServer Cluster 1600 with GPFS, SG24-6954-00
Contents
An Oracle9i RAC implementation over GPFS offers these benefits.
It enables a file system database implementation.
In the past, only RAW devices (logical volumes) were supported under HACMP/ESCRM for the implementation of Oracle9i RAC or Oracle Parallel Server (OPS).
However, using a file system implementation as storage for database files greatly simplifies systems administrator and database administrator tasks (mainly by using the AUTOEXTEND attribute for the tablespaces), as well as other system administration tasks (export, log archiving, backup, and so on). Thus, for many customers, a file system database implementation is preferred.
Unlike most UNIX file systems, which are designed for a single server environment, GPFS allows parallel applications to simultaneously access the same files from any node in the GPFS nodeset. The shared access GPFS is capable of holding the database files, control files, and redo log files required by Oracle9i RAC. It satisfies the Oracle9i RAC shared disk requirement.
It provides RAC with striping I/O performance.
GPFS is designed to provide high performance by striping I/O across the storage subsystem. GPFS stripes the data across all disks that are part of a file system.
It provides direct I/O with close to RAW device performance.
When RAW devices are used, the Oracle bypasses the file buffer cache and accesses the logical volume directly. The RAW logical volume I/O eliminates the file system overhead and the need for AIX virtual memory tuning. This enables Oracle to allocate a larger buffer cache for caching more data in a more direct way. For I/-intensive applications, the use of RAW devices for Oracle datafiles generally results in performance improvement.
For RAC implementation using GPFS, Oracle opens GPFS with direct I/O (thus bypassing the file cache) and transfers data directly from disk into the user space buffer.
Considering the direct I/O feature, performance estimates for running RAC on GPFS can be very close to the performance of Oracle using RAW devices.
It increases data availability and failure survivability.
GPFS also provides high availability through logging and replication. From the cluster perspective, if the cluster loses a node, as long as the GPFS quorum is met, the GPFS file system availability on the rest of the nodes will not be affected.
Special Notices
This material has not been submitted to any formal IBM test and is published AS IS. It has not been the subject of rigorous review. IBM assumes no responsibility for its accuracy or completeness. The use of this information or the implementation of any of these techniques is a client responsibility and depends upon the client's ability to evaluate and integrate them into the client's operational environment.