Conference ulysse::rdb_vms_competition

    SORRY!! 	My mail location is USHS08 not USH08
    				Sparky

    Do you know if the report writer is based on Ally?
    
    Richard

    I had heard that it was, but after contacting oracle people they
    CLAIM, that although they did by Ally they evaluated it and discarded
    it.  I talked to one of the team members that did the evaluation.
    He said they decided it wouldn't fit into the overall direction
    they were going.  He did beleive it was a good product although.
    
    	This report writer looks like a Macintosh interface.  Although
    it just looks like it now, it is not bit mapped, the componants
    are all there, pull down menus, currently accessable only with cursor
    keys or single letter commands.  Oracle claims it will be bit mapped
    for use on the Workstations soon.  They provide several boiler plate
    type reports that they claim will cover about 90% of all reports
    so rather than develop a new report you just edit the boiler plates.
    I haven't used it much, very little.  Oracle also claims it was
    designed by their largest users and not by the oracle people, that
    they went to the users and said here is a rough idea, now what do
    you want. I'll post more as I find out.
    
    By the way, when speaking to customers please don't mention my name
    or location, I have a pretty good relationship with some of the
    Oracle people here.  I know it's hard to say to a customer, I know
    a site, without giving some credibility, but I would hate for the
    customer here to loose their Beta sight and relations with Oracle.
    
    				Thanks Sparky

    I did a little rather unscientific benchmark between Oracle V5.1
    and V6.0.19.  The two databases were up as single node databases,
    the 8700 was on PM, on an 8250 with 8 meg of memory.  The databases
    are identical.  I was the only user on the entire system.  I tried
    a few items.  Keep in mind that this is a single user case, where
    Oracle claims the performance of V6 will be realized with many users.
    However, I beleive it does show that a system doing mostly queries
    will realize very little benifit.  The results are:
    
    (CR just to keep the suspence up)
    With a table of 7,132 records 125 bytes avg length, 14 columns no indexes
    
                                                    V5.1          V6.0.19
                                            elapsed   cpu     elapsed     cpu
    Select Count(*) from table;            00:12.63 00:07.58  00:09.48 00:05.45
    select * from table where ...not found 00:23.48 00:16.74  00:19.14 00:14.75
    create index on key                    00:42.98 00:30.64  01:44.71 01:16.07
    select * from table where ...not found 00:03.52 00:03.43  00:01.52 00:01.49
    select * from table where ...found     00:01.79 00:01.71  00:01.71 00:01.56
    drop index on key                      00:03.05 00:01.53  00:01.96 00:01.87
    export data to disk                    05:33.04 05:06.26  07:01.04 06:34.05
    update all names to null               02:09.82 01:03.89  04:10.01 03:36.16
    drop table                             00:05.83 00:03.22  00:39.42 00:24.86
    import data from disk to database      06:13.21 05:30.08  09:12.00 07:52.94
  
    Anything to do with the data dictionary is painfully slow.  Importing
    our full database in V5 takes about 5 hours complete with index
    build.  It took over 12 hours in V6.  I can't give an details on
    the size or anything.  While searches without indexes are somewhat
    better, this is not something done consistantly.  Searching on items
    that exist with indexes are almost identical.  The updating all
    names to null was a surprise, I tried it twice, same results, no
    answer for that.  That should have been better because the Journaling
    now only writes the bytes that change, not the whole 2048 block
    that one byte was changed in.  All in All I say the only real
    performance increase will be in the use of the TPS PL/SQL.
    They are supposed to be benchmarking that against V6 without TPS
    and V5 here next month, I will try to get the results up. 
    
    
    BTW the director here has been invited to NY for press conferance
    reguarding TPS.  It will aparrently take place the 18-jul-88.  Note
    V6 without TPS will not be available for several months.  This is
    a marketing ploy to get customers who want V6 to buy the extra priced
    TPS package.  Our first beta copy didn't have the TPS PL/SQL.

    Sparky,
    
    Does V6.0 execute under VMS/5.0?  Do you know when ORACLE will execute
    under 5.0?
    
    Regards,
    Nik
    

    re: -1
    
    Local Oracle guys couldn't/wouldn't say about v6 but I have personally
    got V5.1.22 running under VMS V5.0 here on a microVax II.
    
    - Chas
    

    Just got back from being out of town,  We still don't have V5.*
    VMS here, so I don't know if V6 will run on it or not, but I do
    know of another account running V5 oracle on V5 VMS.  I will 
    post when we V5 VMS and they put V6 oracle on it.
    
    		Sparky

    
    
    Sparky,
    
    I have a ton of questions if you happen to run across the answers:
    1) Is the data dictionary really active, metadata changes are reflected
       immediately in the physical schema and affected programs are
       cross-referenced? Can the data dictionary be expanded by the
       user to include his own user interfaces and/or entities. Does
       the dictionary info need to be replicated on each node that uses
       it?
    2) I hear that distributed is not any better, ie no two phase commit
       so only queries, no updates, can be distributed. 
    3) Are there any specific features that Oracle can point to for
       increased performance.
    4) Do they have a decent backup system yet? On-line? with concurrent
       updates? Incremental backup?
    5) What administration facilities require exclusive access of the
       database? Have they put more features on-line?
    6) What is the new version going to cost?
    
    
    Thanks, 
    
    Jamey Nordby

    re: .2
      Hello. I don't follow this note file but someone brought this
    note to my attention. I was one of the original ALLY developers
    and know for a fact the REPORT WRITER is indeed based on ALLY.
    Two of my best friends are heading the development team for the
    REPORT WRITER. ORACLE has gone to a lot of trouble to keep this
    a secret so don't tell anyone.
      BTW. Even though I don't care much for ORACLE, they are one of
    the major success stories of this decade.

    re .1
    >   (Yes you can recover in a cluster environment now)
    Could some pls describe their solution a bit further ?
    - Are there any chances that it's a 'less than 100% solution' 
    	or have reached the level of Rdb on this issue ?
    - Do they use our lock-manager ?
    - Does the cluster-AIJ support require usage of their new TPX ? 
       
    rgds, Henrik