Conference ulysse::rdb_vms_competition

    I just spoke to a project manager of a large software firm that
    has numerous products layered on top of Oracle.  They have been
    a field test site for Oracle V6/TPS for quite some time.  He told
    me that the latest release of V6/TPS is 6.0.25 but he has not received
    it yet.  He added that Oracle was having performance problems in
    certain application environments (didn't know which) and that V6/TPS
    was being held from production until the problems could be found.
    His latest word is that V6/TPS was "planned" for production in Feb
    89 but doubted if it would be ready.  
    
    Last fall the release cycles were fast and furious and we saw the
    variant number (the last two digits) change rapidly and new field
    test kits release every few weeks.  Now that the release cycle has
    stretched to over 2 months, it may mean that they're really having
    trouble finding and fixing bugs.
    
    Could it be that the window of opportunity is beginning to open
    for Rdb ??

    A customer in Swindon, UK, has been testing Oracle V6/TPS and has
    discovered that in their application, with more than 4 (four) users 
    performance                          
    is WORSE than v5. Oracle have blamed this on a routine called SETVBUF
    in our C runtime library, the symptom of which is that when you
    call this routine to be given a buffer of, eg, 8192 bytes, it gives
    you 512 whether you like it or not. This apparently slows their
    forms processing no end. Anyway, C development chaps are fixing
    this buglet, and we shall see whether this helps Oracle out of their
    hole.                                
                                         
    When asked directly by me in front of the customer 'will this fix
    your performance issue', the Oracle technical chap sweated a lot
    and refused to make any guarantee of any kind. So it remains to
    be seen whether they're really in trouble and are using the 'C'
    bug to buy time, or whether that really is the answer. When I hear any
    more, I'll keep you informed.           
                                            
    PDB                                     
                                            
    P.S. if it really is a VMS problem, why didn't their own internal
    testing show this up? Do they do any?   
                                            

    Well, we apparently fixed our bug in the C RTL and it has (so my
    spies tell me) made NO DIFFERENCE AT ALL.
                                           
    Oracle are now sweating even more profusely; it would appear that
    the technical people and even the salesmen have started zooming
    off to the states in a desparate bid to save the situation. 
                                           
    True to form, Oracle have blamed us all along the way, and their
    accusations are apparently becoming wilder and wilder. Fortunately, 
    this is beginning to wear rather thin with the customer, who is,
    by all acounts, more or less frothing at the mouth over it, and
    wants heads on platters. 
                                                          
    I'm rather torn here; I don't like to see one of our customers in
    the mire like this (they've got two and a half years of work invested
    in Oracle; they're too far in to back out now), but on the other
    hand it does this old heart no end of good to see Oracle squirm
    on a knife made by their own foul hand.
    
    I post more from the front when I get it (if you're interested)_.
    
    Cheers
    
    PDB
                            

    I talked with my Oracle contact today and my customer who has decided
    to implement with Oracle V6/TPS on April 1.  They tell me that Oracle
    6.0.26 will be the "Production" release and is now scheduled for
    Feb 28 (this year).  My customer received 6.0.25 today and we are
    rerunning our benchmark on their recently delivered 8840 tonight.

    Also, TPS (Transaction Processing Subsystem) has now been renamed
    to TPO (Transaction Processing Option).  This is interesting since
    their V6/TPS (oops TPO) manuals have already gone to press.  Wonder
    how much they'll have to eat (if any) in manual costs !!
    
    The "Production" release WILL NOT contain support for VAXclusters.
    My customer has been told April, but my Oracle contact told me they
    haven't even began testing the recently inserted code.  He added
    that the pressure was so great to get V6 out the door that cluster
    code was not included ON PURPOSE.  His understanding is that VAXcluster
    support will be included in V6.1 which has no release date so far.
        
    Other items ... in the next few weeks Oracle will announce SQL*forms
    V3.  It will include the PL/SQL.  He mentioned that SQL*forms "needed
    the functionality" of PL/SQL ... whatever that means.  

    
    The customer has given us (informally) some results of the testing
    that they have done with Oracle V6/TPO(S).
    
    System is an 8530. Actual response times are not given (and probably
    wouldn't be meaningful anyway without knowing what the application
    does) but in terms of a base value, x seconds.
           
    		30 users		        4 users  
    V5 Oracle   90% CPU used, x sec response    20% CPU, x sec resp.       
    V6/TPO     100% CPU used, 2x sec response   40% CPU, 0.9x sec resp.
    
    One good point (possibly the only good point) was that the database
    conversion went well (database is 2gb).
    
    So it would appear that V6 can improve your response times slightly
    at the expense of vastly increased CPU usage. If you haven't got
    spare CPU, it will degrade your response times very badly.
    
    The SETVBUF routine was complete red-herring, as Oracle modified
    their code not to use it and it still didn't make any difference
    at all.
    
    Cheers
    
    PDB
    

    Talked with my customer yesterday.  They recently loaded 6.0.25
    with the thought that the index corruption problem had been fixed.
    However, THERE IS STILL A PROBLEM WITH 6.0.25 that CORRUPTS INDEXES.
    
    Oracle is well aware of this bug and tells the customer that "they
    haven't had time to test the fix" which tells me they haven't fixed
    it yet.  The latest thought is that 6.0.26 MAY be the production
    release but Oracle has not decided yet.  No word either if the index
    corruption problem will be fixed in 6.0.26.
    
    -- gerry

    For the latest - see CHEFS::VIA_FORUM note 181.0

    Quotes from internal note given by Oracle to a customer:
          
    "Even with the gains listed above, the CPU  utilisation of ORACLE
    version 6 will always be greater than version 5. This is a fact
    of life."
          
    "It is also a fact that version 6 makes better use of large cpu
    machines and the upgrading of hardware will show clear improvements"
          
    So, if I chose to be cynical (which I do) I could interpret this
    to mean
          
    "Load version 6, and if your performance gets worse (although we
    said with v6 it would get better), buy a bigger machine and you
    might get back what you lost with v6"
                                        
    Or am I too critical?               
                                        
    I trust you will note that this came from from an internal note,
    and therefore not quote it directly directly to customers or mention
    its source.                         
                                        
    Cheers                              
                                        
    PDB                                 
                                        

    Talked with my customer yesterday. They are getting ready to load
    6.0.26.3 which is supposed to be production.  I have verified this
    with my other Oracle user as well.  We should know early next week
    if the index corruption problem still exists.  My customer also
    has HP system running HPUX.  I'm trying to find out if they're using
    V6 there or not cause they're having lots of trouble with their
    HP systems.  Will report as soon as I find out.

    Talked with my customer today.  They were promised 6.0.26 last
    Thursday.  Haven't got it yet.  
    
    FYI  They're using Oracle V5.1 on HPUX.  When I asked them about V6 on
    HPUX, he laughed and said, "... as far as I know, V6 is not available
    on any system except VAX...".
    
    -- gerry 

    Customer I referred to has abandoned a full benchmark with Oracle
    V6, as Oracle couldn't guarantee them a stable product. Odd, since
    as I read some of these other notes it's already out in the field.
                                       
    Oracle have to all intents and purposes given up. Their position
    apparently is 'it goes as fast as it goes, and as for what we led
    you to believe you'd get with v6, well, you aren't going to get
    it'. The customer is beginning to mutter darkly about ACMS and Rdb.
    
    I think we might encourage him quietly.
    
    Cheers 
       
    PDB

    My customer who has been beta test for many months has the following
    versions running:
    
    	Oracle 	V6.0.26.3.0 - Production
    	PL/SQL	V1.0.17.0.0 - Alpha

    It is important to note that "production" is appended to the DBMS
    since the previous version was labeled "limited production".  I
    will find out this week if the index corruption bug has been fixed.