| United States-English |
|
|
|
![]() |
Release Notes for HP-UX 10.30: HP 9000 Computers > Chapter 5 Major Changes for HP-UX 10.10Large File Systems |
|
For 10.10, the maximum size of HFS and JFS file systems has been increased from 4 Gb to 128 Gb. (The maximum size of files continues to be 2 Gb.) Large file systems are supported by HP-UX, including file system administration commands (HFS and JFS), SAM, SD, LVM, accounting commands, quota commands, and backup commands. These interfaces, as in previous releases, allow you to create, modify, and recover file systems. All of the command interfaces remain the same; there are no new options for large file systems. The only difference is the parameter values may now be larger to accommodate file systems greater than 4 Gb. Large file systems can be exported through NFS to other clients, who are be on any other supported version of HP-UX. NFS clients do not have to be on 10.10 to read/write an exported file system. For example, a 10.10 server can export a 20 Gb file system to 9.x clients through NFS and each of the clients can read/write to the 20 Gb file system. Non-HP NFS clients can also access an LFS on an HP server. Increasing the size of a small file system to a large file system is supported through the standard file system extension command extendfs(1M). Using extendfs for JFS is new for HP-UX 10.10. An alternative to extendfs is to back up the data, remake the file system, and restore the data. All file systems built with earlier versions of HP-UX are fully supported in 10.10. There is no conversion required of pre-10.10 file systems to be functional in 10.10. It is not possible to use large file systems for root or boot disks because there are limitations of 2 Gb or 4 Gb (depending on the processor) in the IODC. Applications that access devices directly will continue to have a 4 Gb limit. There is no interface for large devices. The implementation is for large file systems only and is supported through the standard set of administration commands mentioned above. It is possible for users to access file systems directly through their respective device files. The contents can be copied with cp(1) or displayed with od(1). This usage is rare and will continue to be limited to 4 Gb. An exception to this is dd(1), which will support large file devices. For 10.10, there should be no performance degradation. Building and recovering (using fsck) a very large HFS file system will take proportionately longer than a smaller system. For a JFS file system, the scaling is proportional to the log size, not the file system size. Random access performance on both file systems is not impacted. The dcopy command has not been enhanced to support large file systems, but will continue to operate on small file systems. All other commands and interfaces remain intact. fsirand supports the lseek() block modifications for large file systems. If fsirand detects a large file system, all lseek() calls must use block or byte offsets. Previous versions of fsirand will not work properly on a large file system. The current 10.10 version of fsirand should not be run on a previous release. |
||||||||||||||||||||||||||||
|
|||||||||||||||