VERITAS File System 3.5 (HP OnlineJFS/JFS 3.5) Release Notes

New and Changed Features

Chapter 1

New and Changed Features

The VERITAS File System3.5 (HP OnlineJFS/JFS 3.5) release contains the following

new and changed features:

• New and/or Enhanced Tunable Parameters

— hsm_write_prealloc - Preallocates uninitialized space in a managed region of a

ﬁle to improve performance while using the hierarchical storage management

(HSM) applications with the base/full VxFS 3.5.

— read_ahead - Detects the complex read-ahead patterns in addition to sequential

reads.

— write_throttle - Alters the number of dirty pages generated before ﬂushing to

disk, to take advantage of new disk storage technology.

See the vxtunefs (1M) manual page for speciﬁc information.

• VxFS Command Enhancements

The vxdump command now enables backups at the ﬁle or directory level. The

vxrestore command now bypasses the system buffer cache when restoring data

(this enhancement is available only in the full VxFS 3.5 version). This allows a

system's buffer cache to remain intact while doing an extensive restore operation,

which typically improves system performance. The vxrestore -c option preserves

backwards compatibility by allowing vxrestore to use the buffer cache.

Additionally, the vxfsconvert command has been enhanced to allow updating the

disk layout version to version 4 on a non-mounted ﬁle system. The

logiosize=size

option was added to the mount command to enhance the performance of storage

devices that employ a read-modify-write feature. To obtain the maximum

performance from such devices, intent log writes can be done in 1024, 2048, or 4096

bytes.

• New I/O Error Handling Policy

The ioerror option, which has been added to the mount command, provides four

different ways to handle system I/O errors. These policies have been implemented on

the base/full VxFS 3.5 to cater to the evolving storage technology functionality for

which a single method of I/O error processing is no longer sufﬁcient. One major

behavior change is to disable a ﬁle system instead of marking inodes as bad on disk.

This will prevent inadvertent data loss.

See the mount (1M) manual page for more information.

• New Default Intent Log Mode

To increase performance, the mount command delaylog option replaces the log

option as the default for base/full VxFS 3.5 ﬁle systems.

See the mount (1M) manual page for more information.

• New Default File System Block Size

The default block size has been set at 1024 bytes for all base/full VxFS 3.5 ﬁle

systems, no matter how large the ﬁle system you create. Previously, the default block

size was more with larger ﬁle systems. You can still select a larger block size using

the mkfs bsize option.