Difference between HFS vs. VxFS

[Debbie Fleith]

I understand that the difference between HFS and VxFS filesystems is that HFS is non-journaled, and VxFS is journaled.

What does that really mean? Is there any reason why you wouldn't want to create a new filesystem as VxFS under HP-UX 11.0 for use for a database application?

[Rob Smith · Accepted solution]

Hi, there are a number of reasons why you would want to use vxfs as opposed to hfs. Vxfs writes all transactions to a log and then flushes them to disk at intervals. If there is a crash you only have to replay the log, fsck does this for you, to return the filesystem to a stable state. If you have a large hfs file system it can take hours to repair it.

Also, vxfs allocates inodes on the fly whereas hfs has a fixed number of inodes. If you run out of indes you have to rebuild the file system and specify more inodes.

See: http://docs.hp.com/hpux/onlinedocs/B2355-90672/B2355-90672.html for more information. Hope this helps.

Rob

[James R. Ferguson]

Hi Debbie:

"Journaled File Systems" ('JFS', a.k.a. 'vxfs') are the standard type with 11.0. JFS filesytstems can be extended or shrunk dynamically without having to unmount them (with the Online-JFS component). JFS filesystems also provide a variety of mount options which when used appropriately can boost filesystem performance (see: 'man 1m mount_vxfs').

Following a crash, the time to verify the integrity of a filesystem (and to repair it) is greatly reduced for JFS ones over the older HFS filesystem. This is because JFS filesystems keep an "intent log" for replaying the changes made to a filesystem. Thus the entire structure doesn't need to be verified and/or repaired.

...JRF...

[Jared Westgate_1]

Hello,

One great feature of VxFS is the OnlineJFS product. It is a separately licensed product, but I don't think I could live without it.

OnlineJFS lets you do things such as resize filesystems while mounted. You can also defrag a FS. I would consider it a must for a production database environment that requires significant uptime.

Hope this helps,

Jared

[A. Clay Stephenson]

Hi Debbie,

If you are using OnlineJFS for databases you have another option which will result in a noticeable performance increase (I've observed 18%):

use the vxfs -o convosync=direct,mincache=direct,nodatainlog mount options.

Use these for the data and indexes and the normal options for archive/redo logs. The above options bypass the unix file buffers and only use the buffers in the SGA. This gives you all the performance on raw disk while still using conventional cooked files. This also means that you can reduce the unix buffer size and increase the SGA size.

Regards, Clay