3.5.1 Matrix Server Administration Guide

Chapter 7: Configure PSFS Filesystems 107

Persist: This option causes the filesystem to be remounted automatically

when the server is rebooted and is enabled by default. If you do not want

the filesystem to be remounted automatically, remove the checkmark.

Create Directory: If you want Matrix Server to create the directory

mountpoint on each server where the filesystem is to be mounted, click

Create Directory.

Select any other mount options for the filesystem.

Async or Sync. Async, the default, allows either asychronous or

synchronous I/O to the filesystem. Sync allows only synchronous I/O.

Dev or NoDev. Interpret (or do not interpret) character or block special

devices on the filesystem. Dev is the default.

Exec or NoExec. Permit (or do not permit) the execution of binaries on the

mounted filesystem. Exec is the default. Noexec can be used on a system

that has filesystems containing binaries for other architectures.

SUID or NOSUID. Allow (or do not allow) set-user-id bits and set-

group-id bits to take effect. SUID is the default.

Std. File Optimized or DB Optimized.The Std. File Optimized option is

the default and optimizes file access for standard files. The DB Optimized

option is provided with the Matrix Server Database Option product and

is intended primarily for database files. This option bypasses filesystem

I/O buffering, allowing disk transfers to occur directly in application

buffers. In database server terms, this means the I/O will only be buffered

in the address space of the database server processes. For example, in the

case of Oracle9i, I/O will be buffered in the SGA or PGA. This eliminates

the “double-buffering” overhead associated with traditional filesystems.

Buffering directly in database server buffers also improves CPU

utilization because I/O buffers do not need to be copied from Kernel

space to database server buffers. Also, filesystem file data operations are

not serialized, allowing multiple reads and writes at the same time to the

same file from multiple processes. This allows applications to manage

their own file data coherency (e.g., databases) to operate more efficiently.

(The coherency of metadata, or structural, operations continues to be

managed by the PSFS filesystem.)