[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference ulysse::rdb_vms_competition

Title:DEC Rdb against the World
Moderator:HERON::GODFRIND
Created:Fri Jun 12 1987
Last Modified:Thu Feb 23 1995
Last Successful Update:Fri Jun 06 1997
Number of topics:1348
Total number of notes:5438

239.0. "backup ADABAS database" by HGOVC::SHIRLEYCHAN () Thu Nov 03 1988 12:40

	One of our customer is evaluating ADABAS and some other DBMS
    products with Rdb. I heard from them that backing up ADABAS database
    requires unloading the data to a sequential file before using the
    VMS BACKUP command. The database is marked as no backup when built.
    Also the database can't be unload directly to tape. When they want
    to restore the database, they have to build the database firsst. Are 
    all these true?
    
    Thanks for any comment on ADABAS.
    
    
    Shirley
T.RTitleUserPersonal
Name
DateLines
239.1You should see ADABAS code, German comments!COOKIE::JANORDBYThu Nov 03 1988 17:2917
    
    I am not as familiar with the VMS version of ADABAS as I should
    be, but on the IBM side ADABAS can
    1) Unload in its internal compressed format either directly to tape
    or to disk.
    2) Backup the database while it is being updated (3 or 4 years now)
    3) can optionally unload in decompressed format into a sequential
    file. (Part of the load of a sequential file involved data compression)
    4) Loading a file involved building the relation for that file and
    the associated indexes. Restoring a file is usually not done from
    an unload, rather a backup which does not need to rebuild any part
    of the database, it simply places an old copy where it used to go.
    (if the origninal space is not available, anerror occurs)
    5) I assume that the VMS version is at least along the same designs
    as the IBM version, since it was actually a port to VMS.
    
    Jamey
239.2Same Physical Space ?HGOVC::SIMONYTCHANI Come From Hong KongFri Nov 04 1988 09:4611
    4) Loading a file involved building the relation for that file and
    the associated indexes. Restoring a file is usually not done from
    an unload, rather a backup which does not need to rebuild any part
    of the database, it simply places an old copy where it used to go.
->  (if the origninal space is not available, anerror occurs)

    Does it mean that the same physical space as that of the original
    copy must be available in order to restore the backup?

    						Simon Chan

239.3DifferencesBROKE::BOOTHWhat am I?...An Oracle?Fri Nov 04 1988 17:2017
    Adabas is written by two totally seperate development teams in Germany.
    The IBM team writes in Assembler. The Digital team writes in Macro.
    
    Consequently, the two products are never exactly in sync. There
    are subtle differences in most areas, major ones elsewhere.
    
    When I was in the field, the Adabas/VMS customers complained that
    it was misery to restart the database after a system or database
    problem. Hours were involved in bringing up the database again.
    
    If online backup is available on the VAX, it is brand-new. If you
    have other Adabas customers, talk to them about these questions.
    Check out everything you can. The typical situation with Adabas/Natural
    is that the user/developers love it, while the systems managers
    hate it, because of the work they must do maintaining it.
    
    ---- Michael Booth
239.4yupCOOKIE::JANORDBYFri Nov 04 1988 18:1111
    re: .2
    In the IBM version, a backup will backup a physical copy of a file
    or the entire database. If a file is to be restored, the physical
    space that it previously occupied must be available. An unload is
    a different mechanism (and slower). But an unload can be used as
    input to a load that is independent of previous location.
    
    Again, you must double check with customers to be sure this is the
    same with the VMS version of ADABAS.
    
    Jamey
239.5Adabas (VMS) Backup detailsWELMT2::DENTFri Oct 27 1989 17:2425
    Adabas (VMS) - all versions up to and including V1.5 use information
    about the physical block numbers on disk to speed up access to data
    with QIO. In addition chunks of the database are aligned on cylinder
    boundaries on disk to speed up performance by reducing disk head
    movement. 
    
    As Adabas uses physical disk blocks it is not possible to do a VMS
    image backup and restore (Physical backup is ok) as the image backup
    is almost certain to restore the data to different disk blocks.
    
    The way to take a backup of the whole database or one or more files
    in the database is with the Adabas utility ADABCK. This may be run
    with the database active and may write data to 1 or 2 tapes or a
    disk file. The disk file may then be copied to tape with VMS Backup.
    
    As ADABCK takes advantage of knowing the physical layout of the
    database and only backs up the parts of the database used it is
    quicker than VMS Backup. 
    
    With the release of Adabas (VMS) V2 (approx end 89) the database
    may be created as a "easy to use" or "high performance" database.
    If "easy to use" then the restrictions on cylinder alignment and
    contiguous files are removed and VMS image backup may be used.
    
    Ian Dent
239.6TWOBOB::CLEARYA deviant having fun..."Sun Nov 05 1989 07:248
    If he only difference between `easy to use' and `high performance' is
    the use of logical QIO's instead of physical QIO's, then I'd be very
    surprised if the performance hit is noticable let alone measurable for
    most applications.  It's always been a source of amazement to me that
    ADABAS did this.  But then I've heard they write it all in assembler so
    what else would you expect...
    
    -mark