The ADSM server does bypass JFS buffering using O_SYNC on the open of
the JFS file.  There is no problem using JFS for the ADSM server
database recovery log and storage pool volumes, and this is, in fact, the
method that I suggest to customers most often.

Mike Kaczmarski
ADSM Development

Pierre Verpeaux writes:
 > I think you should use raw LV. It is the only way to make ADSM aware
 > of what is really written on the disk. The JFS buffering can improve
 > performances, but in case of a system crash, your database can be
 > inconsistent. As i understand it, the JFS log concern only the filesystem
 > structure and not the file's content.
 >
 > If there is a problem with database corruption when the server is not
 > correctly shut down, that is a bug in ADSM which must be corrected.
 > A database manager must ensure the integrity of the database but for
 > this, it needs to know the real state of the database (what is the content
 > of the disk ? ).
 >
 > If there is a way to bypass the JFS buffering, i would be thery happy to
 > know it as i need it for performance reason.
 >
 > Pierre Verpeaux