Conference 7.286::digital

      I saw part of the event, via DVN broadcast - Palmer's & Ellison's 
    presentations. Themes that were emphasized include: 64 bits, cluster 
    reliability over mainframe, Oracle in-memory database, Alpha processor 
    speed, ease of programming, price, and Ellison's goofy high-pitched laughs.

      The message was clear - target IBM's mainframe customers. 

      Blurb that did not compute:  Ellison shows slides of Oracle 7's improved 
    performance on in-memory Alpha system. He states that Tandem holds the 
    TpmC record but KNOWS that if Oracle ran the TpmC benchmark, it would win.  
    Hmmm, another 35mm projection.  He can't seem to break the habit.
    
      If I was an IBM mainframe customer, I would want to see proof of a 
    migration solution for all my applications before I considered Digital
    and Alpha systems.

    I attended the event in NYC...the actual announcement, the
    press/analyst Q&A and the product showcase.  Overall, went very
    well.  Major themes are speed/performance, capacity, reliability.
    Our offerings are TurboLaser platforms, 64-bit UNIX operating system,
    64-bit compilers, 64-bit databases, SMP and CLusters,database expertise 
    centers, solutions expertise centers, applications/solutions 
    partnerships,etc.  Bob Palmer & Larry Ellison did a very good job 
    highlighting what this breakthrough technology can deliver in terms of 
    solutions to real customer problems.  Pauline Nist pulled a great 
    recovery from the teleprompter problems and discussed the actual box and
    architecture.  Clear customer targets are our installed base, Oracle
    installed base, mainframe downsising customers.  Enrico Pesatori
    spoke of the changes made in Digital, and some of the new marketing
    and revenue generation programs and customer support initiatives.
    
    Although the major message was performance, VLM/VLDB capacity, etc.,
    I thought we got some major messages across.
    
    Mark
    

    Was Pauline Nist really wearing one black and one red shoe, or did it
    just look that way on the DVN?
    
    Anyhow, the portion I saw of the announcement looked real good, even
    with the couple teleprompter snafus.  The ad agency did a nice job!  I
    only had time to catch a portion of the announcement, but it was nice
    to be able to see it after all the work engineering has put over the
    years.
    
    /Charlotte

    Pauline shoes were half black and half red so if she turned it looked
    like one red and one black but facing front you saw them correctly.
    BTW, Pauline did a great comeback and did a good presentation.
    
    I thought the annoucement went quite well in NY. My customer that
    attended told me afterward it was the best Digital annoucement he has
    ever seen. 
    
    I thought the Oracle annoucement with us helped bill the credibility
    to our annoucement. Ellison does a great job.

    It appears that investors weren't impressed initially.  DEC stock
    reached 42 1/2 yesterday by the annoucement time,  closed at 41 1/2
    and opened at 40 1/4 this morning.  They have since recovered to
    41 5/8 as of 11:19 this morning.
    
    Robin
    
    
Your requested quote...courtesy of PC Quote and SpaceCom Systems

DIGITAL EQUIPMENT CORP 

Symbol   : DEC    
Exchange : New York 

===[ Last Trade ]===============================

   Price : 41  5/8   +
   Size  : 2000 
   Time  : 11:19 

===[ Daily Record ]=============================

   Open  : 40  1/4  
   High  : 41  5/8  
   Low   : 40  1/4  
   Volume: 799600 


Quote information is believed correct but is not warranted
or guaranteed by PC Quote or SpaceCom Systems

    I wouldn't expect the stock to go up because of the announcement,
    at least not afterwards.  "Buy on the rumor, sell on the news"
    
    				Steve

    
    
    I did not see any article in today's WSJ or the SJ Merc -
    of course I could have missed them too! I'll be appalled if
    the annc did not even merit a line of press beyond the Globe.
    
    BTW I heard the Globe (!!!!) gave it a positive spin. Would
    anyone be kind enough to post that article here?
    
    Thanks.
    		-Ashikin
    
    
    

    WBNW Newsradio 590 out of Boston ("Bloomburg Radio Network") gave the
    announcement some good airplay yesterday afternoon and evening.  They
    even had a pretty good BP soundbite (about 10 seconds).  It was the
    first time I've heard that from a DEC Exec.!  Only problem was their
    statement that this was (and I'm paraphrasing here) "...one of the
    fastest Oracle DB configs..."  I'm no techie, but I thought this Oracle
    7 on Turbolaser Alpha was supposed to be THE fastest!!  Or, is that
    just our hype?
    
    

    Technically I am missing something.  Regular windows is 16bits;
    Windows32 is 32 bits.  Win32 will be faster than Win16.  An application
    which can take advantage of 32 bit addressing will be faster because it
    (or the OS behind its back) does NOT have to do the segmenting.  But
    nobody has claimed that Win32 delivers twice or 10 times or heaven
    forbid, 300 times performance of Win16.
    
    So how come 64 bit Oracle is delivering 200-300 times better performance
    than Oracle 32?  Is it possible that the performance gains are coming
    from using the flat address space on the *memory* vs having to use
    flat addressing space on the *disk*?  Aren't we comparing the memory
    speeds vs disk speeds?
    
    The real performance advantages are because instaling 14GB of memory
    can now be done "cost-effectively".  Most of the systems on this planet
    today don't even have 16GB of disk space! 
    
    Technically, without having the 64bit capability there is no way to put
    that much memory in a single system.  This is a big hammer approach
    to the performance but it does work.
    
    The next step should be to go for 2GB of secondary cache for our next
    generation SuperTurboPropFan Laser system!
    
    - Vikas

    re: .9
    
    The performance increases discussed during the announcement in NY by
    Larry Ellison were comparisons to what PREVIOUSLY was the fastest box
    that ran Oracle. Previous to this, the DEC 7000 running Digital UNIX
    was the Oracle record holder. Now it is the turbolaser. I don't have
    the exact numbers, but they were something like this:
    
    							    8400
    				DEC 7000	8400	Performance
    
    Time to read a big file in:   time		time	3x faster
    Time to do ad hoc query:	  time		time	20x faster
    Time to do a 3-way join:	  time		time	20x faster
    Time to do a 4-way join:	  time		time	69x faster
    Time to do a 5-way join:	  time 		time	278x faster
    
    Oracle, as stated earlier in this conference, doesn't do TPC/C
    benchmarks. They prefer to 'benchmark' a system doing something a 'real
    customer cares about'. The above is an example of how they prefer to
    handle system speed comparisons. 
    
    Another example used took a real customers application. Oracle ran it
    on an 8400 with VLM capability turned off. From a 'hot start' (data
    already loaded into memory), it took 1 hour 27 minutes to run the application and get
    the results. Using the same machine with VLM turned on, it took only 27
    minutes. 
    
    So, it depends on what you're doing with a TurboLaser how much increase
    in speed you get. However, it is very clear that it is CLEARLY the
    fastest thing that runs Oracle today, and it is clear that as queries
    get more complex, the advantages of the TurboLaser increase (Larry
    Ellison's observation, not mine.)
    
    

    Re: .9
    
    Yes, the comparison is the in-memory system vs. a disk system.  What
    makes this interesting is that on 32-bit systems you can't address
    enough data to be interesting to most large database users.  Digital
    UNIX is the only (I think) system that can do that.  Then you add the
    ability to stuff the cabinet with enough RAM, and you've got the
    "killer app".  It solves a real problem for real customers, as
    outlandish as it may seem.
    
    				Steve

    .9 as stated in the few previously - the point was made by Palmer and
    Ellison that this is not magic, the performance gain is simply by using
    memory instaed of disk.  The figure given by Palmer was that memory is
    10,000 times faster than disk.  So a 200x improvement in performance is
    quite easily explainable.  
    
    I think a lot of people in the industry have been waiting for the day
    when memory was cheap enough and the system could address enough of it
    that you could stuff your whole application in memory and run it there. 
    That was the point of the announcement - we're there now, no magic.

RE: .11


	Just considering how far we've come in compute power, disk space,
	and memory in the past 10 years is mind-boggling.. 10 years from now, 
	what has been announce today won't be that outlandish. We'll
	probably have that 14GB and 300Mhz in our laptops. :)

						mike

One thing to remember:

The biggest performance increases are for "query only," which does not require
disk access when the entire database in loaded into main memory.  Transaction
processing, in which records must be updated on disk, still runs faster on the 
new AlphaServers, but at a smaller multiple of two to five times.

    re: .10
    
    Thank you for recording the numbers!  They look like what I remember.
    
    But I think you got the column headers wrong.  Wasn't it TURBOLASER
    using 32-bit ORACLE vs. TURBOLASER using 64-bit?  
    
    I thought Larry Ellison specifically made the point that the
    comparisons were without any hardware changes, just showing off the
    capabilities of 64-bit-Oracle; and that if he had also included
    hardware changes (e.g. comparing to an SGI or Sun system in 32 bit mode
    to Turbolaser in 64-bit) the factors would have been even bigger.
    
    	/john

I attended the event yesterday in London, accompanying a customer.   
While the numbers were impressive, and while it became clear why users 
might want 64 bits, I have to say that the event was an embarrassment.

The tone was set by the so-called comedian they hired as MC.  He did two 
routines at the beginning as a warm-up for the audience that met mostly 
with stunned silence.  The material was offensive and irrelevant; I 
wanted to hide under my chair.

Aside from Theo Wegbrans and the UK product marketing manager, we had a 
marketing guy from Computer Associates (to whom the general reaction was 
"so what?") and Oracle's IT director for the UK.  The latter might have 
been all right, had he talked about how 64-bits made it possible for him 
to run his business, but he didn't:  he just tried to give a 
techno-marketing pitch, and it didn't come off.

The piece de resistance was a Turbolaser appearing in smoke, with lots 
of spots focused on it.  Unfortunately, there was so much smoke, that 
you couldn't see it.

Sigh.

    re:  Oracle performance
    
    32-bit Oracle pulls data off disk into memory in 2K blocks.  64-bit
    Oracle pulls data off disk in 32K blocks, also known as BOB (Big
    Oracle Blocks).  Getting the memory & pipes filled faster helps the
    performance.  (ps...the 32-bit Oracle is designed to move data in
    8K blocks, but for various reasons, they only were able to effectively
    and safely use 2K blocks).  
    
    Msrk
    

That's one piece I didn't understand.  If Big Oracle Blocks (BOB) at 32KB are
so much better, why is it that existing users, who have the option today of
using 8KB blocks instead of the default of 2KB, on the whole choose _not_ to
use the larger blocks (as Mr. Ellison pointed out multiple times)?

    Mr. Ellison pointed out that current users could use 8KB blocks but
    choose the 2KB blocks instead because there aren't enough 8KB blocks to
    make it worth while doing so.   There must be a tradeoff between faster
    transfer size and memory wastage.
    
    Not all the data in a block is useful.  Some of it is irrelevant.
    4 2KB blocks probably contain more interesting data that 1 8KB block. 
    Enough more to make up for the slower transfer rate.
    
    32KB probably have an even smaller density of information but there are
    so many of them that it doesn't matter.
    
                                August

    -1. is right. The BOB (Big Oracle Blocks) would possibly increase the
    contention. Particularly, in the OLTP environment, several different
    transactions accessing/updating different rows in the same block will
    contend for the block. Random update would run slower.
    Furthermore, unused data in the big block unnecessarily reduces
    available I/O bandwidth.
    The BOB will help the Decision Support applications or Very Large Rows
    and Multimedia applications. I'm not sure it would benefit the OLTP.

    
    Note that Oracle Rdb has VLM too on Digital UNIX and does not
    require a special version for that.
    
    _Luc ( Oracle now )

     re -1  ?? what was that about Oracle Rdb on Digital UNIX?
    	    I thought Rdb is on OpenVMS, and that it may be avail
       	    on Unix (ours only I assume), but that was before it 
    	    was sold.
    
    	    I'm curious, so would you pls elaborate?
    
    Few back --  Obviously there is a need to explain more clearly
    	how Oracle is exploiting the 64-bitness via VLM and VLDB. Also
    	obviously there are tradeoffs to be considered, as Ngugen had
    	pointed out earlier. Whether VLM or VLDB works for your situation
    	depends quite entirely on understanding what the application does.
    	DSS and Datawarehousing MAY be more suitable for a VLM
        implementation than a TPC-C type OLTP application.
    
    What we Digital folks need to appreciate first and understand is that
    this is probably the first time there is a coordinated and synchronized
    effort to fundamentally match our hardware features & capabilities to
    a software engine (a partner's that is); it so happens that Oracle is 
    the most ready to do this than the Sybase or Informix, but we are 
    working on that front even more so than before.
    
    	-Ashikin
     

    Is there a video available yet, for customer presentations and such,
    have tried to find in VTX but our system was changed around and I
    can't find any order numbers.
    
    Thanks.

    <---- is there a video being made ??
    
    I believe there are efforts under way to edit the announcement day
    presentations down a bit and produce a customer-consumable video for
    field use. I'll try to find out who the best source is for you to
    contact. Feel free to contact me if you don't hear something in a few
    days.
    
    /se
    

    Yes, Oracle Rdb on Digital UNIX is available today
    and it supports VLM databases via 64 bit addressing.
    
    Moreover, Oracle Rdb will support 64 bit addressing on
    OpenVMS as soon as OpenVMS provides that support.
    
    

    Oracle Rdb v6.1 on Digital UNIX v3.2 is released and available.
    Thru it's feature called 'global buffers' it has all the functionality
    to do VLM since v4.1 . We were simply waiting for the hardware and
    operating system that can handle massive memory caches to exploit this
    at a maximum. No need to invent a new marketing term 'VLM' for that,
    which probably says more about Digital versus Oracle marketing and
    long term 'core competencies' decisions :-) .
    
    _Luc

IBM Angry over new Digital range
--------------------------------
By Louise Kehoe in San Francisco for the Financial Times, April 15th


    IBM and Digital Equipment, arch rivals in the computer industry, are
    waging a war of words over conflicting claims about the performance
    of products which could be critical to the companies' future.

    The confrontation has been sparked by Digital's introduction,
    earlier this week, of a new generation of computers based on its
    latest Alpha microprocessor chips, which are widely acknowledged to
    be the fastest devices of their kind.

    Digital claims that the new machines are much faster and cheaper
    than IBM's mainframes.  IBM counters that Digital's performance
    claims are "misleading and unsubstantiated".

    IBM is expected to launch a campaign next week with newspaper
    advertisements to rebut Digital's statements and thos of Mr. Larry
    Ellison, chief executive of Oracle, the leading dtabase software
    company which jointly developed software  for the new Alpha
    computers with Digital.


    At the Digital product introduction, Mr Ellison said that Oracle
    customer tests indicated that Digital's new servers are 18 times
    faster than the most powerful mainframe computer, at one tenth of
    the cost.  He predicted that Digital would "gain tremendouss market
    share" at the expense of IBM's mainframes.  Digital said that it
    supports Mr Ellison's statements.

    Some independent market researchers agree Digital's new Alpha
    servers represent a breakthrough in computer technology.  "They are
    64-bit chips, which can address a huge amount of memory.  That
    should give Digital a significant performance advantage over all its
    competitors," said Mr Tom Willmott, vice-president of the Aberdeen
    Group, a market research firm.

    IBM executives complain that Digital has failed to use the
    appropriate industry tests in its comparisons.  These tests measure
    the number of transactions per minute that can be handled by a
    computer and are audited by the indepenndent Transaction Processing
    Performance Council.  IBM says that Digital has yet to submit its
    new computers for testing, so the comparisons it has made are
    unfair.

    Mr Willmott said the TPC benchmarks were the "only legitimate
    measure" of performance.  He notes, however, that IBM has also
    failed to submit its mainframe computers to TPC tests.

    The Digital Alpha servers challenge IBM's core mainframe
    businesswhich has contributed to the company's return to
    profitability in 1994 after it lost $16bn in the previous three
    years.

    Yet the success of the new Alpha servers is also critical for
    Digital.  Struggling to regain its momentum after four years of
    heavy losses, Digital has spent hundreds of millions of dollars on
    developing chips and built new plants to produce them.

    What a nice score .. in particular, I like the following excerpt:

|    Mr Willmott said the TPC benchmarks were the "only legitimate
|    measure" of performance.  He notes, however, that IBM has also
|    failed to submit its mainframe computers to TPC tests.

    >Per

    It's great to be compared with IBM again!  And, all that free ink!
    Love it!
    
    Steve

    Just spell our name right :-)
    

    "It's not true, and even if it were true, it's not fair!", says IBM's
    performance VP "Stinky" Henderson.  Stinky then added "Nanny, nanny,
    boo, boo" and other pertinent comments before being given a "time-out".
    
    							Tex

    Tex, you are too funny! (Honest, too!!)  If IBM had any way to compete
    with the new Alpha Servers, they wouldn't be whining about tests and
    comparisons.  Of course, as previously noted, all this "bad press" from
    IBM can't help but build even more interest in these terrific new
    systems.
    
    IMHO.
    
    M.
    

From conference 
            <<< NOMAHS::DISK$NOMAHS1:[NOTES$LIBRARY]RDB_60.NOTE;1 >>>

================================================================================
Note 9.46                       Support Articles                        46 of 46
NOVA::SMITHI "Don't understate or underestimate Rdb!" 326 lines   2-MAY-1995 13:18
--------------------------------------------------------------------------------
                     -< Oracle Rdb and Very Large Memory >-

Please share with any and all customers!




                     INTRODUCTION to ORACLE VLM


   WHAT IS ORACLE VLM?
   Oracle VLM (Very Large Memory provides the ability to support
   system configuration with greater than four gigbytes of RAM.
   Most UNIX systems are limited to at most four gigbytes or RAM,
   bases on the memory addressing limitations of 32-bit architectures.  
   Oracle 7 and Oracle Rdb running on Digital's Alpha UNIX systems 
   take advantage of a pure 64-bit implementation to support memory 
   configurations above four gigbytes.  Oracle is architected to support 
   RAM configurations of hundreds of gigbytes, well beyond the 
   hardware limitations of today's memory technology.

               
   WHAT ARE THE BENEFITS OF VLM?
   The primary benefit of VLM is PERFORMANCE, because many
   operations can now run at memory speed instead of disk speed.  With 
   larger amounts of data in memory, the database issues many fewer calls 
   to disk, thus eliminating the delay of disk I/O.  Preliminary results 
   indicate performance improvements of greater than 100% for operations 
   such as multi-way joins.

   VLM also provides the advantage of SCALABILITY, to support larger 
   numbers of users and larger amounts of data.  They system does not need
   to swap data in and out of memory to process all of the transaction 
   requests and can more effectively accommodate requests for larger 
   amounts of data.


   WHAT NEW SOLUTION DOES VLM PROVIDE?
   With VLM, Oracle can support new types of applications that simply 
   were not possible in the past, because of performance limitations.  
   VLM promises significant performance improvements for transactions 
   that access large amounts of data, stored in many different database 
   tables, with very large table sizes, supporting large numbers of users,
   and with complex query types.  Without VLM, any one of these 
   characteristics can provide significant performance bottlenecks.

   VLM opens new opportunities for organizations looking to manage
   ever-increasing amounts of data, particularly in industries such as health 
   care, financial services, insurance, government, communications, 
   transportation, retail, and manufacturing.


   HOW DOES VLM RELATE TO ORACLE'S ENTERPRISE SOLUTION PROGRAM?
   Oracle's Enterprise Solution Program (ESP) is a corporate initiative for
   customers who would like to achieve the benefits of open systems for 
   their business applications.  The Enterprise Solutions Program consists
   of a number of different components, including the Oracle7 products,
   technologies, consulting services, support services, and applications
   such as SAS, SAP, CICS, and Oracle Applications.

   For many applications that currently run on proprietary mainframe
   systems, Oracle7 running on 64-bit Digital VLM configurations
   provides the most viable open systems alternative.  VLM provides a
   key enabling technology for ESP, with a new level of scalability for
   enterprise-wide applications supporting large numbers of users and large
   amounts of data.


   WHAT IS THE DIFFERENCE BETWEEN VLM AND VLDB?
   VLDB or Very Large Database, refers to the ability to support large amounts
   of data stored on disk.  While no precise definition exists to
   distinguish a "very large" database, this term frequently applies to
   databases in the 300 gigabyte range, extending up to multiple terabytes.

   VLM differs from VLDB in that VLM refers to data stored in memory,
   while VLDB refers to data stored on disk.  VLM is a major technology
   breakthrough for supporting very large databases.  VLM allows a larger
   portion of the database to reside in memory.  As a result, MORE users
   can access MORE data FASTER, with less time spent moving data to
   and from disk.  Without VLM, applications must move data back and 
   forth between RAM, and disk to accommodate the limitations of two
   gigabytes of memory.  This I/O process is very time consuming and
   limits performance.


                        PRICING/AVAILABILITY

   WHEN WILL VLM BE AVAILABLE?
   Support for VLM is available now on both Oracle7 (V7.1.6) and
   Oracle Rdb (Release 6.1) for Digital UNIX platforms.

  
   ON WHAT PLATFORMS DOES ORACLE SUPPORT VLM?
   Currently, Oracle7 and Oracle Rdb support VLM on Digital's 7000 and
   8000 series systems running Digital UNIX (formerly Digital OSF/1).
   In the future, Oracle7 and Oracle Rdb will support VLM and DEC 2100
   AlphaServers, when these systems can support large memory hardware 
   configurations.

   The Digital/Oracle solution provides the only true 64-bit solution, with
   (1) a processor architected for 64-bits, (2) an operating system that was
   designed for 64-bits from the ground up (without having to backward 
   design for an installed base running 32-bits), and (3) a database
   optimized for 64-bits.


   DOES ORACLE PLAN TO SUPPORT VLM ON OpenVMS?
   Oracle is working to support of VLM technology in both Oracle7 and
   Oracle Rdb for OpenVMS Alpha systems.  Although the company is
   not prepared to make a product availability announcement at this time,
   Oracle expects that such support will appear first on Oracle Rdb shortly
   after 64-bit support is available from Digital on OpenVMS V7.0
   around the end of 1995.

   
   ON WHAT DATABASES DOES DIGITAL SUPPORT VLM?
   Digital supports VLM on Oracle7 and Oracle Rdb.  Oracle is the first
   database vendor to take advantage of Digital's 64-bit architecture to
   support VLM database configurations.


   HOW IS ORACLE VLM PRICED?
   Oracle VLM is priced as an option to the database; however, as a
   special promotion, Oracle is offering VLM at no charge with the
   current releases of Oracle7 (7.1.6) and Oracle Rdb (6.1).

                           APPLICATION ENABLEMENT

   WHAT ARE THE PERQUISITES FOR VLM?
   VLM is best suited to systems that are architected for pure 64-bits,
   including the hardware, operating system, file system, and database
   system.  Without a pure 64-bit architecture, systems cannot effectively
   achieve the benefits of VLM technology.  Many solutions will 
   implement a partial 64-bit operating system must often accommodate an
   existing installed base running with 32-bit applications on 32-bit
   hardware.  These limitations can affect their ability to implement a
   suitable VLM solution.

 
   WHAT TYPES OF APPLICATIONS BENEFIT FROM VLM?
   VLM can benefit both query-intensive (DSS) and read-write (OLTP)
   transactions.  For DSS, VLM provides particular advantage for index
   builds, full table scans, and hoc queries, and multi-way joins.  For
   OLTP, VLM provides the ability to support very large tables, large
   amounts of data, and large numbers of users.

   For customers who are evaluating alternatives to improve database
   performance, VLM provides a viable alternative to time consuming
   optimization tasks such as redesigning physical and logical database
   structures.  Customers can achieve immediate performance gains by
   simply adding memory beyond their existing 4GB limit.


   WHAT SIZE OF DATABASE WORKS BEST WITH VLM?
   VLM can benefit any size database, not just very large databases.  First,
   with VLM, Oracle can store ENTIRE databases in memory.  For larger
   databases, VLM allows larger portions of the database (those portions 
   most frequently accessed by users) to reside in memory.

   Second, VLM also allows intermediate operations to reside in memory,
   such as creating temporary tables.  In essence, VLM provides a larger
   workspace for any database operation.  So, even with database of one
   or two gigabytes, VLM can provide significant advantages for complex
   transactions, such as multi-way joins, that can require several
   intermediate operations.


   HOW HAS ORACLE IMPLEMENTED VLM SUPPORT IN ORACLE7?
   Oracle VLM uses the same base Oracle7 product, which is designed to
   support 64-bit environments.  As a result, Oracle VLM running on
   Digital 64-bit systems offers the same functionality as the standard
   Oracle7 product.  Oracle has optimized performance for 64-bits with
   Large System Global Area (LSGA) and Big Oracle Blocks (BOB).

   Large System Global Area refers to an Oracle database buffer cache of
   greater than four gigabytes.  The buffer cache is the portion of memory
   where the database system stores database records.  A larger SGA allows
   the database to store proportionately more data in RAM.  LSGA takes
   advantage of Oracle's 64-bit architecture, combined with the ability of
   the operating system to support 64-bit memory addressing.  In
   conjunction with LSGA, Oracle also plans to support Large Sort Work
   Areas in VLM configurations.

   Big Oracle Blocks provide the ability to support larger I/O transfers
   between memory and disk.  BOB complements large SGA
   configurations, because BOB allows the system to move data faster
   between memory and disk.  With VLM configurations, system performance
   depends directly on the ability of the system to move
   database blocks into the SGA as efficiently as possible.  Without the
   benefits of improved data transfer, VLM, performance can decline.
   

   HOW HAS ORACLE IMPLEMENTED VLM SUPPORT IN ORACLE Rdb?
   Rdb is engineered to support very large global buffer caches (similar to
   LSGA), to fully utilize the 64-bit memory addressing and the large memory
   capacity of Digital's AlphaServers.  Oracle Rdb also supports
   large I/O transfers between memory and disk (similar to BOB).

   
   DOES VLM REQUIRE ANY CHANGES TO EXISTING APPLICATIONS?
   VLM is transparent to all Oracle-based applications, other than
   delivering marked improvements in performance.

                            CONFIGURATION

   HOW DO I CONFIGURE ORACLE7 AND ORACLE Rdb FOR VLM?
   Support for 64-bit VLM configurations is included in the standard
   Oracle7 and Oracle Rdb products installed on my Digital UNIX system
   supporting large memory configurations.


   MEMORY IS EXPENSIVE RELATIVE TO ADDING ADDITIONAL CPU's.  WHEN DO I USE
   SMP VERSUS MORE MEMORY TO IMPROVE PERFORMANCE?
   First, SMP and VLM are complementary.  Both will provide incremental
   throughput and performance.  Together, they provide a
   performance multiplier affect.  The specific performance benefit will of
   course be dependent on the application.


   HOW DO I KNOW WHAT THE IDEAL MIX OF SMP AND MEMORY IS FOR MY APPLICATION?
   Oracle believe that all application environments will gain cost
   effective performance improvements utilizing directly addressable, large
   memory configurations.  Oracle Services has consultants who are
   trained and available for further assistance.

                           
                     RELIABLE, SCALABLE MANAGEABLE

   HOW DOES VLM AFFECT, DATABASE RELIABILITY?
   Oracle7 and Oracle Rdb on Digital VLM configurations provide the
   same degree of high reliability and data integrity as in non-VLM
   configurations.  Database integrity is completely isolated from the
   volatility of the RAM.  The amount of data in memory has no
   effect on the reliability of the system.

   Oracle7 with VLM uses the standard Oracle7 redo logs to ensure that
   updates and changes to the database are written to disk before a transaction
   is committed.  Oracle has architected the redo logs to
   provide a fast and efficient way to write data to disk via single or
   multiple streams of I/O.

   As another example, Rdb implements an algorithm for accelerated
   commitment through electron disks (ACE), making it possible to
   significantly enhance journal I/O by using a very small, but fast solid
   state disk.


   WHAT SYSTEM MANAGEMENT TOOLS SUPPORT VLM?
   Polycenter/Netview provides a full suite of tools for system and data
   management across the network and the enterprise from a single
   console.  For instance, Oracle7 is fully integrated with
   Polycenter/Netview.  Oracle7 provides a SNMP-compliant subagent
   used to create an Oracle map and a series of Oracle specific graphs.


   HOW DO I BACKUP A DATABASE ON A VLM SYSTEM?
   system administration tasks such as backup and restore are no different
   for VLM systems than any other Oracle7 or Oracle Rdb configuration.
   Today, DEC NSR (Legato) is fully integrated with Polycenter/Netview
   and provides a complete set of back-up and restore capabilities.  Later
   this year Oracle will offer Oracle Parallel Backup and Restore (OPBR)
   as well as planned support for Open Vision's Open V*NetBackup, for
   those customers who require higher performance backups.


   HOW DOES VLM RELATE TO SMP AND CLUSTERS?
   With support for VLM and clusters, Oracle and Digital provide
   scalability through a three-tiered approach: processors (SMP), memory
   (VLM), and systems/nodes (clusters).  This strategy provides the 
   requisite foundation for supporting enterprise-wide solutions on an open 
   system platform.

   SMP allows customers to add compute power through additional
   processors, all accessing a common memory pool.  For SMP systems,
   VLM provides a larger buffer cache, shared by all of the processors.
   Algorithms for navigating and manipulating in-memory locks and 
   buffers will continue to remain efficient even as memory expands from
   tens of megabytes to ten of hundred of gigabytes.

   Likewise, with clusters, customers can network multiple systems to
   work in parallel.  The systems themselves can each support large
   memory configurations, with internode communications managed by a
   high speed interconnect such as Digital's upcoming Memory Channel
   architecture.

 
   DOES VLM WORK WITH ORACLE PARALLEL SERVER?
   Yes.  Oracle Parallel Server will support VLM configurations on
   Digital AlphaServer cluster expected later this year.  Digital's Memory
   Channel technology will provide a high performance interconnect for
   multiple AlphaServer systems each configured for VLM.  The result is
   a clustered system supporting tens, or even hundreds, or gigbytes of
   RAM, all running Oracle7 on Digital UNIX.  Oracle is currently
   investigating Oracle Parallel Server and Oracle Rdb support for
   OpenVMS with Memory Channel.


                              SERVICE AND SUPPORT

   WHAT SUPPORT SERVICES ARE AVAILABLE FOR VLM?
   Oracle provides the same support services for VLM as for all other
   Oracle products.  Customers have the choice of Bronze, Silver, or Gold
   support levels, depending on their particular application requirements.
   For customer who have implemented enterprise-wide applications,
   Oracle Gold provides 7x24 mission Critical Support levels for the operating
   system and offers Mission Critical Support levels for the operating system
   and hardware, which complement Oracle Gold.  Oracle and Digital are
   actively working to develop joint support offerings for customers as
   part of the Enterprise Solutions Program.


   WHAT CONSULTING SERVICES ARE AVAILABLE FOR VLM?
   Oracle has established a team of VLM experts in the company's
   Advanced Technologies Group.  The ATG specializes in the design,
   architecture, and deployment of enterprise-wide client/server systems.
   VLM represents only one of many levels of specialization within the
   group.  As such, Oracle has the expertise to provide you with a
   complete solution for your enterprise.

   For customers interested in Oracle VLM, Oracle strongly recommends
   working with the experts in the ATG to realize the benefits of this
   technology.  These experts have already implemented solutions using
   VLM and have the hands-on-experience needed to make your application
   a success.  The group has established best practices around
   implementing VLM solutions quickly and effectively within an 
   enterprise.